Email alert
2022 Vol. 44, No. 2
column
- cover
- contents
- Special Topic on Brain Computer Interface and Human Computer Interaction
- Pattern Recognition and Intelligent Information Processing
- Cryptography and Information Security
- Wireless Communication and Internet of Things
- Radar, Sonar and Array Singal Procossing
- Electromagnetic field and Electromagnetic Wave
- Satellite Navigation
Display Method:
2022, 44(2): 1-4.
Abstract:
2022, 44(2): 415-423.
doi: 10.11999/JEIT211337
Abstract:
The game Brain-Computer Interface (game-BCI) is a novel interactive mode that the user can control and interact with the game directly by recognizing brain signals. It not only provides a new way of game interaction for healthy people, but also provides a new way of rehabilitation for the disabled. Compared with the other invasive BCIs, BCI based on scalp ElectroEncephaloGram (EEG) has a wider application prospects because there are the advantages of non-invasiveness, high time resolution, low cost, and good portability. The game-BCI technology based on scalp EEG is summarized and is divided into active, reactive, passive, and hybrid paradigms according to control signal types. Then the control strategies and application scenarios of different types of game BCI are introduced. The classification algorithms of EEG signals commonly used in game BCI are compared and analyzed. Finally, the current problems and future development directions in this filed are discussed.
The game Brain-Computer Interface (game-BCI) is a novel interactive mode that the user can control and interact with the game directly by recognizing brain signals. It not only provides a new way of game interaction for healthy people, but also provides a new way of rehabilitation for the disabled. Compared with the other invasive BCIs, BCI based on scalp ElectroEncephaloGram (EEG) has a wider application prospects because there are the advantages of non-invasiveness, high time resolution, low cost, and good portability. The game-BCI technology based on scalp EEG is summarized and is divided into active, reactive, passive, and hybrid paradigms according to control signal types. Then the control strategies and application scenarios of different types of game BCI are introduced. The classification algorithms of EEG signals commonly used in game BCI are compared and analyzed. Finally, the current problems and future development directions in this filed are discussed.
2022, 44(2): 424-436.
doi: 10.11999/JEIT210534
Abstract:
Emotion is a subjective feeling to internal or external events with positive and negative meanings, and plays an important role in human's daily life. Emotion decoding aims to automatically discriminate different emotional states by decoding physiological signals. Through estimating emotion changes, it helps to solve the practical problems in the clinical applications and develop more friendly human-interaction systems. Nowadays,functional Magnetic Resonance Imaging (fMRI)based emotion decoding is one of the most commonly used methods for deepening our understanding of the emotion-related brain dynamics and boosting the development of affective intelligence.This paper introduces current research progress, applications and main problems in the field of fMRI-based emotion decoding technique, covering experimental design, data acquisition, existing emotion-related fMRI dataset, data processing, feature extraction, and emotional pattern learning and classification.
Emotion is a subjective feeling to internal or external events with positive and negative meanings, and plays an important role in human's daily life. Emotion decoding aims to automatically discriminate different emotional states by decoding physiological signals. Through estimating emotion changes, it helps to solve the practical problems in the clinical applications and develop more friendly human-interaction systems. Nowadays,functional Magnetic Resonance Imaging (fMRI)based emotion decoding is one of the most commonly used methods for deepening our understanding of the emotion-related brain dynamics and boosting the development of affective intelligence.This paper introduces current research progress, applications and main problems in the field of fMRI-based emotion decoding technique, covering experimental design, data acquisition, existing emotion-related fMRI dataset, data processing, feature extraction, and emotional pattern learning and classification.
2022, 44(2): 437-445.
doi: 10.11999/JEIT211001
Abstract:
In order to improve the initiative of hemiplegic patients and system anti-interference in the mirror rehabilitation training process, an upper limb mirror control strategy based on adaptive assist-as-needed is proposed. The strategy includes mainly two modules: mirror control and adaptive assist-as-needed control. The mirror module collects the position of the healthy side to calculate the desired position of the affected side, and compares it with the actual position of the affected side to obtain the position deviation; The adaptive assist-as-needed module combines thetraditional impedance control and the method for evaluating the movement state of the affected limb, automatically adjusts the auxiliary force to the affected limb in real time to maximize the active torque of the affected limb. Two experiments are designed to verify the effect of this strategy on patient initiative and system anti-interference. The experimental results show that the use of the proposed method for the mirror rehabilitation training system reduces effectively the average auxiliary force by 56.9%, and the position accuracy of the affected limb that follows the healthy limb is 5.6%, and the proposed method can effectively compensate 89% of the interfering external force. Therefore, the method improves effectively the initiative of the trainer, and at the same time has good anti-interference, and meets the requirements of mirror rehabilitation training.
In order to improve the initiative of hemiplegic patients and system anti-interference in the mirror rehabilitation training process, an upper limb mirror control strategy based on adaptive assist-as-needed is proposed. The strategy includes mainly two modules: mirror control and adaptive assist-as-needed control. The mirror module collects the position of the healthy side to calculate the desired position of the affected side, and compares it with the actual position of the affected side to obtain the position deviation; The adaptive assist-as-needed module combines thetraditional impedance control and the method for evaluating the movement state of the affected limb, automatically adjusts the auxiliary force to the affected limb in real time to maximize the active torque of the affected limb. Two experiments are designed to verify the effect of this strategy on patient initiative and system anti-interference. The experimental results show that the use of the proposed method for the mirror rehabilitation training system reduces effectively the average auxiliary force by 56.9%, and the position accuracy of the affected limb that follows the healthy limb is 5.6%, and the proposed method can effectively compensate 89% of the interfering external force. Therefore, the method improves effectively the initiative of the trainer, and at the same time has good anti-interference, and meets the requirements of mirror rehabilitation training.
2022, 44(2): 446-454.
doi: 10.11999/JEIT210816
Abstract:
A deep learning method for Steady-State Visual Evoked Potential (SSVEP) classification is proposed to solve the problem that phase and frequency information are not fully used in existing deep learning models. First, the proposed model uses complex vectors of fast Fourier transform as input and operates convolution on real and imaginary vectors to learn phase information, and then utilizes the spatial attention module to enhance discriminative frequency information. Next, two-dimensional convolution and max pooling are used to extract further spatial and frequency features. Finally, fully connected layers are utilized to classify. The accuracy of proposed model can reach 81.21% in the case of cross subject, and the accuracy can be further improved to 83.17% by adding the standard sinusoidal signal templates to the training set. The results show that the proposed model achieves better performance than canonical correlation analysis algorithm.
A deep learning method for Steady-State Visual Evoked Potential (SSVEP) classification is proposed to solve the problem that phase and frequency information are not fully used in existing deep learning models. First, the proposed model uses complex vectors of fast Fourier transform as input and operates convolution on real and imaginary vectors to learn phase information, and then utilizes the spatial attention module to enhance discriminative frequency information. Next, two-dimensional convolution and max pooling are used to extract further spatial and frequency features. Finally, fully connected layers are utilized to classify. The accuracy of proposed model can reach 81.21% in the case of cross subject, and the accuracy can be further improved to 83.17% by adding the standard sinusoidal signal templates to the training set. The results show that the proposed model achieves better performance than canonical correlation analysis algorithm.
2022, 44(2): 455-463.
doi: 10.11999/JEIT211040
Abstract:
For the SSVEP-BCI, an ergonomic study is carried out on the effects of stimulation interface element size and spacing on recognition efficiency and user experience. In this experiment, the red squares are used as the stimulation elements. The squares are located on the upper, lower, left and right positions. The independent variables include two factors: size and spacing. Factor 1 (size) is the side length of the square, which is divided into three levels: 100px, 150px, and 200px; Factor 2 (spacing) is the vertical/horizontal distance between the element center and the interface center, which is divided into three levels: 200px/400px, 300px/600px, and 400px/800px. The dependent variables are the completion time and the number of failures of the tasks. Subjective evaluation is carried out after the experiment. Based on ISO 9241 usability standard, Likert 7-point scale is used to score the participants’ satisfaction of the interfaces. The results of the ergonomics experiment show that the element size has a significant impact on the recognition efficiency, the stimulation element with side length of 200px has the highest recognition efficiency, while the element spacing has no impact. The subjective evaluation results show that element spacing has a significant impact on user satisfaction. The compactness (200px/400px) or alienation (400px/800px) of stimulating elements will lead to the decline of satisfaction. The satisfaction of 300px/600px spacing level is the best, while the size has no impact. From the perspective of design ergonomics, it is found that the size and spacing of stimulating interface elements have an impact on the efficiency of SSVEP-BCI system and user satisfaction respectively. The research conclusion has guiding and reference value for standardizing the design of SSVEP-BCI and improving the efficiency of SSVEP-BCI system.
For the SSVEP-BCI, an ergonomic study is carried out on the effects of stimulation interface element size and spacing on recognition efficiency and user experience. In this experiment, the red squares are used as the stimulation elements. The squares are located on the upper, lower, left and right positions. The independent variables include two factors: size and spacing. Factor 1 (size) is the side length of the square, which is divided into three levels: 100px, 150px, and 200px; Factor 2 (spacing) is the vertical/horizontal distance between the element center and the interface center, which is divided into three levels: 200px/400px, 300px/600px, and 400px/800px. The dependent variables are the completion time and the number of failures of the tasks. Subjective evaluation is carried out after the experiment. Based on ISO 9241 usability standard, Likert 7-point scale is used to score the participants’ satisfaction of the interfaces. The results of the ergonomics experiment show that the element size has a significant impact on the recognition efficiency, the stimulation element with side length of 200px has the highest recognition efficiency, while the element spacing has no impact. The subjective evaluation results show that element spacing has a significant impact on user satisfaction. The compactness (200px/400px) or alienation (400px/800px) of stimulating elements will lead to the decline of satisfaction. The satisfaction of 300px/600px spacing level is the best, while the size has no impact. From the perspective of design ergonomics, it is found that the size and spacing of stimulating interface elements have an impact on the efficiency of SSVEP-BCI system and user satisfaction respectively. The research conclusion has guiding and reference value for standardizing the design of SSVEP-BCI and improving the efficiency of SSVEP-BCI system.
2022, 44(2): 464-476.
doi: 10.11999/JEIT210845
Abstract:
EEG (ElectroEncephaloGram) signal is susceptible to various of artifacts due to its low amplitude and poor SNR (Signal-Noise Ratio). Among this noise, the ocular artifacts usually hold higher amplitude and strong randomness which would cause serious distortion on EEG signal, and result in great influence on the subsequent analysis. However, traditional methods fail to locate the artifacts components accurately, leading to the loss of the efficient signal components. In order to solve the above problem, this paper proposes a data-driven based automatically artifact-localization-and-removement method. In this paper, the local density is firstly introduced into ICA (Independent Component Analysis) so as to estimate the adaptive threshold with clustering strategy. This adaptive threshold would be further used to noise localization and removal. Subsequently, this paper compared the performance differences between the proposed method and the traditional methods through simulation and the real resting-state EEG experiments. The results with indexes such as PSNR (Peak Signal-to-Noise Ratio), MSE (Mean Square Error), and MI (Mutual Information) quantitatively verify the significant superiority of the proposed method to other ICA-based ocular artifacts removal strategies through statistical analysis.
EEG (ElectroEncephaloGram) signal is susceptible to various of artifacts due to its low amplitude and poor SNR (Signal-Noise Ratio). Among this noise, the ocular artifacts usually hold higher amplitude and strong randomness which would cause serious distortion on EEG signal, and result in great influence on the subsequent analysis. However, traditional methods fail to locate the artifacts components accurately, leading to the loss of the efficient signal components. In order to solve the above problem, this paper proposes a data-driven based automatically artifact-localization-and-removement method. In this paper, the local density is firstly introduced into ICA (Independent Component Analysis) so as to estimate the adaptive threshold with clustering strategy. This adaptive threshold would be further used to noise localization and removal. Subsequently, this paper compared the performance differences between the proposed method and the traditional methods through simulation and the real resting-state EEG experiments. The results with indexes such as PSNR (Peak Signal-to-Noise Ratio), MSE (Mean Square Error), and MI (Mutual Information) quantitatively verify the significant superiority of the proposed method to other ICA-based ocular artifacts removal strategies through statistical analysis.
2022, 44(2): 477-485.
doi: 10.11999/JEIT210778
Abstract:
In Motor Imagery (MI) based Brain Computer Interface (BCI), more channels of ElectroEncephaloGram (EEG) signal are usually adopted to improve the classification accuracy. But there will be channels containing irrelevant or redundant information about MI tasks, which degenerate the performance improvement of BCI. A Channel Selection method based on Correlation and Sparse Representation (CSR-CS) is proposed for EEG classification. Firstly, the Pearson correlation coefficient of each channel of the training sample is calculated to select the significant channels. Then the filter bank common spatial pattern features of the region where the significant channels are located are extracted and spliced into a dictionary. The number of non-zero sparse coefficients obtained from the dictionary is used to characterize the classification ability of each region, and the significant channels contained in the significant regions are selected as the optimal channels. Finally, the common spatial pattern and support vector machine are employed for feature extraction and classification respectively. In the classification experiments of two categories of MI task with BCI competition III dataset IVa and BCI competition IV dataset I, the average classification accuracy reaches 88.61% and 83.9%, which indicates the effectiveness and robustness of the proposed channel selection method.
In Motor Imagery (MI) based Brain Computer Interface (BCI), more channels of ElectroEncephaloGram (EEG) signal are usually adopted to improve the classification accuracy. But there will be channels containing irrelevant or redundant information about MI tasks, which degenerate the performance improvement of BCI. A Channel Selection method based on Correlation and Sparse Representation (CSR-CS) is proposed for EEG classification. Firstly, the Pearson correlation coefficient of each channel of the training sample is calculated to select the significant channels. Then the filter bank common spatial pattern features of the region where the significant channels are located are extracted and spliced into a dictionary. The number of non-zero sparse coefficients obtained from the dictionary is used to characterize the classification ability of each region, and the significant channels contained in the significant regions are selected as the optimal channels. Finally, the common spatial pattern and support vector machine are employed for feature extraction and classification respectively. In the classification experiments of two categories of MI task with BCI competition III dataset IVa and BCI competition IV dataset I, the average classification accuracy reaches 88.61% and 83.9%, which indicates the effectiveness and robustness of the proposed channel selection method.
2022, 44(2): 486-495.
doi: 10.11999/JEIT210601
Abstract:
In view of the existing human-computer interaction systems, which generally exist in the robot’s emotional response lack of uniqueness and initiative, and low participation degree, satisfaction degree and experience sense of participants, a model of emotional response of robots based on fuzzy cognitive map based on the emotional space of Pleasure-Arousal-Dominance (PAD) is proposed. Firstly, the participants’ interactive input emotion values are obtained, and the participants’ emotional state matrix is obtained by evaluating them. Secondly, considering the personality characteristics and social roles of the robot, the relationship between the current emotional state and the long-term emotional state of the context is established by using the time continuity of fuzzy cognitive map, so as to model the robot emotional response process. Finally, according to the influence of the robot’s personality and social role on the emotional response, the emotional state matrix of the robot is updated and mapped into the continuous emotional space to obtain the robot’s emotional response. The model comparison experiment results show that the proposed model can increase the initiative and uniqueness of the robot emotional response, and effectively improve the satisfaction of the participants, and increase the participants’ sense of experience.
In view of the existing human-computer interaction systems, which generally exist in the robot’s emotional response lack of uniqueness and initiative, and low participation degree, satisfaction degree and experience sense of participants, a model of emotional response of robots based on fuzzy cognitive map based on the emotional space of Pleasure-Arousal-Dominance (PAD) is proposed. Firstly, the participants’ interactive input emotion values are obtained, and the participants’ emotional state matrix is obtained by evaluating them. Secondly, considering the personality characteristics and social roles of the robot, the relationship between the current emotional state and the long-term emotional state of the context is established by using the time continuity of fuzzy cognitive map, so as to model the robot emotional response process. Finally, according to the influence of the robot’s personality and social role on the emotional response, the emotional state matrix of the robot is updated and mapped into the continuous emotional space to obtain the robot’s emotional response. The model comparison experiment results show that the proposed model can increase the initiative and uniqueness of the robot emotional response, and effectively improve the satisfaction of the participants, and increase the participants’ sense of experience.
2022, 44(2): 496-506.
doi: 10.11999/JEIT210465
Abstract:
At present, brain-controlled robotic arms have shown broad application prospects in many fields such as medical rehabilitation, but they also have disadvantages such as poor flexibility and fatigue of users. In view of the above shortcomings, an asynchronous control system based on Steady-State Visual Evoked Potential (SSVEP) in an Augmented Reality (AR) environment is designed. A Filter Bank Canonical Correlation Analysis (FBCCA) is applied to identify 12 targets. A dynamic window based on voting strategy and difference prediction is proposed to adjust the stimulus duration adaptively. The robotic arm is asynchronously controlled by pseudo-key to complete the task of the Jigsaw Puzzle. The experimental results demonstrate that the dynamic window can automatically adjust the length of stimulation according to the state of subjects. The average offline accuracy is (93.11±5.85)％, the average offline ITR is (59.69±8.11) bit·min–1. The average selection time of an online single command is 2.18 s. It can reduce the visual fatigue of the subjects effectively. Each subject can accomplish the puzzle task quickly, which indicates the feasibility of this human-computer interaction method.
At present, brain-controlled robotic arms have shown broad application prospects in many fields such as medical rehabilitation, but they also have disadvantages such as poor flexibility and fatigue of users. In view of the above shortcomings, an asynchronous control system based on Steady-State Visual Evoked Potential (SSVEP) in an Augmented Reality (AR) environment is designed. A Filter Bank Canonical Correlation Analysis (FBCCA) is applied to identify 12 targets. A dynamic window based on voting strategy and difference prediction is proposed to adjust the stimulus duration adaptively. The robotic arm is asynchronously controlled by pseudo-key to complete the task of the Jigsaw Puzzle. The experimental results demonstrate that the dynamic window can automatically adjust the length of stimulation according to the state of subjects. The average offline accuracy is (93.11±5.85)％, the average offline ITR is (59.69±8.11) bit·min–1. The average selection time of an online single command is 2.18 s. It can reduce the visual fatigue of the subjects effectively. Each subject can accomplish the puzzle task quickly, which indicates the feasibility of this human-computer interaction method.
2022, 44(2): 507-513.
doi: 10.11999/JEIT201097
Abstract:
The use of appropriate abnormal data cleaning algorithms in the Internet of Things (IoT) can greatly improve data quality. Statistical methods or clustering methods are utilized to clean anomalies in Spatio-temporal data. However, these methods require additional prior knowledge, which will incur additional computational overhead for the sink node. In this paper, in line with the low-rank sparse matrix decomposition model, a fast anomaly cleaning algorithm based on a deep neural network is proposed to solve the Spatio-temporal data cleaning problem in IoT. Both the Spatio-temporal correlation of sensing data and the abnormal values' sparsity are considered in an optimization problem. The Iterative Shrinkage-Thresholding Algorithm (ISTA) is used to solve it. Then the ISTA is unfolded into a fixed-length deep neural network. The real-world dataset’s experimental results show that the proposed method can automatically update the thresholds faster and more accurately than the traditional ISTA.
The use of appropriate abnormal data cleaning algorithms in the Internet of Things (IoT) can greatly improve data quality. Statistical methods or clustering methods are utilized to clean anomalies in Spatio-temporal data. However, these methods require additional prior knowledge, which will incur additional computational overhead for the sink node. In this paper, in line with the low-rank sparse matrix decomposition model, a fast anomaly cleaning algorithm based on a deep neural network is proposed to solve the Spatio-temporal data cleaning problem in IoT. Both the Spatio-temporal correlation of sensing data and the abnormal values' sparsity are considered in an optimization problem. The Iterative Shrinkage-Thresholding Algorithm (ISTA) is used to solve it. Then the ISTA is unfolded into a fixed-length deep neural network. The real-world dataset’s experimental results show that the proposed method can automatically update the thresholds faster and more accurately than the traditional ISTA.
2022, 44(2): 514-522.
doi: 10.11999/JEIT210015
Abstract:
An ultrasound image segmentation method of thyroid nodules based on the improved u-net network is proposed in this paper, in order to solve the problem of changeable size of thyroid nodules and difficulty in segmentation due to edge blur of thyroid nodules in the ultrasound image. Firstly, the image is downscaled to extract the features through an encoder path with a residual structure and a multi-scale convolution structure. Secondly, the long skip connection with an attention module is used to maintain the edge contour of characteristic tensor. Finally, the segmentation result is obtained by a decoder path with a residual structure and a multi-scale convolution structure. The experimental results show that with the method proposed in this paper, the average segmentation Dice value reaches 0.7822. It indicates that this method has better segmentation performance than the traditional U-Net method.
An ultrasound image segmentation method of thyroid nodules based on the improved u-net network is proposed in this paper, in order to solve the problem of changeable size of thyroid nodules and difficulty in segmentation due to edge blur of thyroid nodules in the ultrasound image. Firstly, the image is downscaled to extract the features through an encoder path with a residual structure and a multi-scale convolution structure. Secondly, the long skip connection with an attention module is used to maintain the edge contour of characteristic tensor. Finally, the segmentation result is obtained by a decoder path with a residual structure and a multi-scale convolution structure. The experimental results show that with the method proposed in this paper, the average segmentation Dice value reaches 0.7822. It indicates that this method has better segmentation performance than the traditional U-Net method.
2022, 44(2): 523-533.
doi: 10.11999/JEIT201025
Abstract:
In order to solve the problem that the target tracking algorithm based on the discriminant spatial regularization term has a high mistracking rate under the interference of occlusion, rotation and other factors, an adaptive spatial and anomaly target tracking is proposed. Firstly, an adaptive spatial regularization term is constructed in the objective function, which not only alleviates the influence of boundary effect, but also improves the resolution of the filter between the target and the background region. Secondly, the verification score is calculated according to the response value of each frame, and the reliability and abnormality of the tracking results are analyzed. Finally, the updating rate of target model and response model is dynamically evaluated. A large number of experimental results show that the target tracking algorithm based on adaptive spatial anomaly can deal with background blur, shape change and other abnormal situations well, and has robust tracking performance.
In order to solve the problem that the target tracking algorithm based on the discriminant spatial regularization term has a high mistracking rate under the interference of occlusion, rotation and other factors, an adaptive spatial and anomaly target tracking is proposed. Firstly, an adaptive spatial regularization term is constructed in the objective function, which not only alleviates the influence of boundary effect, but also improves the resolution of the filter between the target and the background region. Secondly, the verification score is calculated according to the response value of each frame, and the reliability and abnormality of the tracking results are analyzed. Finally, the updating rate of target model and response model is dynamically evaluated. A large number of experimental results show that the target tracking algorithm based on adaptive spatial anomaly can deal with background blur, shape change and other abnormal situations well, and has robust tracking performance.
2022, 44(2): 534-542.
doi: 10.11999/JEIT210097
Abstract:
In this paper, the firing modes and spike frequencies of the fractional-order extended Hindmarsh-Rose(HR) neuronal model under Transcranial Magneto-Acoustical Stimulation (TMAS) are investigated. The TMAS with different parameters generate different alternating current and further have various effect on the firing characteristics of the neuronal model. To address the effect of TMAS on firing characteristics under different ultrasound and magnetic field parameters, the membrane potential curves and bifurcation diagrams are exhibited and analyzed. The results show that the firing mode and spike frequency are strongly dependent on the ultrasonic and magnetic field intensities. It is also found that there is no influence of the ultrasonic frequency on the firing mode, though it changes the firing frequency over a small range. Moreover, compared with the integer-order neuronal model, the fractional-order extended HR neuronal model exhibits more variable firing modes and more complex discharge rhythms. These conclusions reveal the influencing mechanism of TMAS and can be taken as theoretical basis for TMAS experimental and clinical application.
In this paper, the firing modes and spike frequencies of the fractional-order extended Hindmarsh-Rose(HR) neuronal model under Transcranial Magneto-Acoustical Stimulation (TMAS) are investigated. The TMAS with different parameters generate different alternating current and further have various effect on the firing characteristics of the neuronal model. To address the effect of TMAS on firing characteristics under different ultrasound and magnetic field parameters, the membrane potential curves and bifurcation diagrams are exhibited and analyzed. The results show that the firing mode and spike frequency are strongly dependent on the ultrasonic and magnetic field intensities. It is also found that there is no influence of the ultrasonic frequency on the firing mode, though it changes the firing frequency over a small range. Moreover, compared with the integer-order neuronal model, the fractional-order extended HR neuronal model exhibits more variable firing modes and more complex discharge rhythms. These conclusions reveal the influencing mechanism of TMAS and can be taken as theoretical basis for TMAS experimental and clinical application.
2022, 44(2): 543-551.
doi: 10.11999/JEIT210052
Abstract:
Remote sensing images have rich content, and then the features extracted by the general depth model are easily interfered by the complex background. The key features can not be extracted well, and it is difficult to express the spatial information of the image. A deep convolutional neural network based on multi-scale pooling and norm attention mechanism is proposed, which weights adaptively salient features at the channel level and the spatial level. First, in the multi-scale pooling channel attention module, the max pooling of different scales is performed on the feature map of each channel based on spatial pyramid pooling. Next, the feature maps of different sizes are transformed to a uniform size by adaptive average pooling. Thus the salient features of different scales can be paid attention by element-wise addition. Then, in the norm spatial attention module, the pixels corresponding to the same spatial position of each channel are formed into vectors, and the feature map with spatial information is obtained by calculating the L1 norm and L2 norm of the vector group. Finally, the cascaded pooling method is adopted to optimize the high-level features, and the high-level features are used for remote sensing image retrieval. Experiment are conducted on UC Merced data set, AID data set and NWPU-RESISC45 data set. The results show that the proposed attention model improves the retrieval performance by concerning the salient features of different scales and combining the spatial information.
Remote sensing images have rich content, and then the features extracted by the general depth model are easily interfered by the complex background. The key features can not be extracted well, and it is difficult to express the spatial information of the image. A deep convolutional neural network based on multi-scale pooling and norm attention mechanism is proposed, which weights adaptively salient features at the channel level and the spatial level. First, in the multi-scale pooling channel attention module, the max pooling of different scales is performed on the feature map of each channel based on spatial pyramid pooling. Next, the feature maps of different sizes are transformed to a uniform size by adaptive average pooling. Thus the salient features of different scales can be paid attention by element-wise addition. Then, in the norm spatial attention module, the pixels corresponding to the same spatial position of each channel are formed into vectors, and the feature map with spatial information is obtained by calculating the L1 norm and L2 norm of the vector group. Finally, the cascaded pooling method is adopted to optimize the high-level features, and the high-level features are used for remote sensing image retrieval. Experiment are conducted on UC Merced data set, AID data set and NWPU-RESISC45 data set. The results show that the proposed attention model improves the retrieval performance by concerning the salient features of different scales and combining the spatial information.
2022, 44(2): 552-565.
doi: 10.11999/JEIT210011
Abstract:
Considering the problems of low accuracy of personalized recommendation and sensitivity to cold start, low-rank Probabilistic Matrix Factorization recommendation model incorporating Multiple Weighting Factors (MWFPMF) is proposed; The trust network is constructed using a given social network, and the trust between users is calculated using the Page rank algorithm and trust transfer mechanism; The user’s social status is calculated based on Page rank, and the weight of the relationship between users is modified using activity scores and scoring time; Term Frequency-Inverse Document Frequency(TF-IDF) is introduced to take user tags, and the homogeneity between users is characterized by tag similarity; The three factors of trust among users, influence of users’ social status, and user homogeneity are integrated into the low-rank probability matrix decomposition, so that user preferences and activity characteristics are mapped to the same low-rank space, and the user-activity scoring matrix is decomposed. Under the premise of regularization as a constraint, the effective prediction of the lack of user ratings by the low-rank feature matrix is finally completed. The data sets of Douban Beijing and Ciao are used to determine the parameter settings of each module. Through simulation and comparison experiments, it can be seen that this recommendation model obtains higher recommendation model accuracy. Compared with the other five traditional recommendation algorithms, the mean absolute error is reduced by at least 6.58%, and the mean square error is reduced by at least 6.27%, compared with the deep learning advancing algorithm, the recommendation accuracy is almost the same; It has obvious advantages in cold-start user recommendation. Compared with other recommendation algorithms, the average absolute error is reduced by at least 0.89%, and the mean square error is reduced by at least 3.01%.
Considering the problems of low accuracy of personalized recommendation and sensitivity to cold start, low-rank Probabilistic Matrix Factorization recommendation model incorporating Multiple Weighting Factors (MWFPMF) is proposed; The trust network is constructed using a given social network, and the trust between users is calculated using the Page rank algorithm and trust transfer mechanism; The user’s social status is calculated based on Page rank, and the weight of the relationship between users is modified using activity scores and scoring time; Term Frequency-Inverse Document Frequency(TF-IDF) is introduced to take user tags, and the homogeneity between users is characterized by tag similarity; The three factors of trust among users, influence of users’ social status, and user homogeneity are integrated into the low-rank probability matrix decomposition, so that user preferences and activity characteristics are mapped to the same low-rank space, and the user-activity scoring matrix is decomposed. Under the premise of regularization as a constraint, the effective prediction of the lack of user ratings by the low-rank feature matrix is finally completed. The data sets of Douban Beijing and Ciao are used to determine the parameter settings of each module. Through simulation and comparison experiments, it can be seen that this recommendation model obtains higher recommendation model accuracy. Compared with the other five traditional recommendation algorithms, the mean absolute error is reduced by at least 6.58%, and the mean square error is reduced by at least 6.27%, compared with the deep learning advancing algorithm, the recommendation accuracy is almost the same; It has obvious advantages in cold-start user recommendation. Compared with other recommendation algorithms, the average absolute error is reduced by at least 0.89%, and the mean square error is reduced by at least 3.01%.
2022, 44(2): 566-572.
doi: 10.11999/JEIT210114
Abstract:
The performance of number theoretic transformation in lattice-based cryptography is insufficient, and the number theoretic transformation parameters are different. A Random Access Memory (RAM)-based reconfigurable multi-lanes number theoretic transform is proposed. In the design of number theory transformation unit, the multi-lanes architecture is improved on the time decimation operation architecture, and an optimized address allocation method is proposed. The number theory transform unit is implemented on Xilinx artix-7 Field Programmable Gate Array (FPGA) platform. The results show that the resource consumed by the unit is 1744 slices and 16 DSP, and the time to complete a polynomial multiplication is 2.01 μs (n=256), 3.57 μs (n=512), 6.71 μs (n=1024) and 13.43 μs (n=2048). The unit supports reconfigurable configurations of 256～2048 parameters n and 13～32-bit modulus q, and the maximum operating frequency is 232 MHz.
The performance of number theoretic transformation in lattice-based cryptography is insufficient, and the number theoretic transformation parameters are different. A Random Access Memory (RAM)-based reconfigurable multi-lanes number theoretic transform is proposed. In the design of number theory transformation unit, the multi-lanes architecture is improved on the time decimation operation architecture, and an optimized address allocation method is proposed. The number theory transform unit is implemented on Xilinx artix-7 Field Programmable Gate Array (FPGA) platform. The results show that the resource consumed by the unit is 1744 slices and 16 DSP, and the time to complete a polynomial multiplication is 2.01 μs (n=256), 3.57 μs (n=512), 6.71 μs (n=1024) and 13.43 μs (n=2048). The unit supports reconfigurable configurations of 256～2048 parameters n and 13～32-bit modulus q, and the maximum operating frequency is 232 MHz.
2022, 44(2): 573-580.
doi: 10.11999/JEIT201093
Abstract:
In order to improve the restoration effect of Visual Secret Sharing (VSS), a visual multi-secret sharing scheme based on random grid is proposed. By using a multi secret image sharing scheme based on random mesh threshold of cylindrical surface, users can share multiple secret images at one time. This scheme has good robustness, so even part of the share images are tampered, the secret image can still be recovered. At the same time, the number of shares has a positive correlation with the visual quality of the final recovered image. Simulation results show that the proposed multi secret vision sharing scheme is two times better than original single secret vision sharing scheme in terms of pixel recovery accuracy, that is, it increases the number of secret images and improves the recovery accuracy.
In order to improve the restoration effect of Visual Secret Sharing (VSS), a visual multi-secret sharing scheme based on random grid is proposed. By using a multi secret image sharing scheme based on random mesh threshold of cylindrical surface, users can share multiple secret images at one time. This scheme has good robustness, so even part of the share images are tampered, the secret image can still be recovered. At the same time, the number of shares has a positive correlation with the visual quality of the final recovered image. Simulation results show that the proposed multi secret vision sharing scheme is two times better than original single secret vision sharing scheme in terms of pixel recovery accuracy, that is, it increases the number of secret images and improves the recovery accuracy.
2022, 44(2): 581-590.
doi: 10.11999/JEIT210009
Abstract:
Multi-Pattern Matching(MPM) works as a core algorithm of packet processing procedure. In order to improve the performance, an efficient packet segmentation and parallel pattern matching algorithm–DCPM (Distance Comparison Parallel Matching) is proposed based on the Aho-Corasick (AC) algorithm. Comparing with existing solutions, DCPM eliminates the threads’ synchronization overhead and decreases the redundant detection overhead. The DCPM algorithm is evaluated on an eight-core processor server platform. The experimental results show that the performance is largely improved (1.3～3.5 times when processing real-world traffic with 8 threads, compared with existing solutions). Meanwhile, the performance of DCPM is less affected by the proportion of pattern strings in the traffic, the length of pattern strings, as well as the number of states in automata.
Multi-Pattern Matching(MPM) works as a core algorithm of packet processing procedure. In order to improve the performance, an efficient packet segmentation and parallel pattern matching algorithm–DCPM (Distance Comparison Parallel Matching) is proposed based on the Aho-Corasick (AC) algorithm. Comparing with existing solutions, DCPM eliminates the threads’ synchronization overhead and decreases the redundant detection overhead. The DCPM algorithm is evaluated on an eight-core processor server platform. The experimental results show that the performance is largely improved (1.3～3.5 times when processing real-world traffic with 8 threads, compared with existing solutions). Meanwhile, the performance of DCPM is less affected by the proportion of pattern strings in the traffic, the length of pattern strings, as well as the number of states in automata.
2022, 44(2): 591-601.
doi: 10.11999/JEIT210683
Abstract:
Splitting authentication codes are an important method to study authentication codes with arbitration. Splitting authentication codes have a higher utilization rate of encoding rules than non-splitting authentication codes. Splitting authentication codes are constructed through group divisible design in this article. Firstly, a theorem for constructing splitting authentication codes is given. The theorem uses Group Divisible Design (GDD) to construct a splitting-GDD, and then a splitting-Balanced Incomplete Block Design (BIBD) by splitting-GDD is constructed, and then a splitting authentication code is obtained; Secondly, the feasibility of constructing splitting authentication codes through GDD under the conditions given in this article is verified. Then a splitting design is given and a splitting authentication codes based on GDD is constructed; Thirdly, the number of sources, the number of encoding rules, the number of messages of the splitting authentication code, the impersonation attack probability and the substitution attack probability are calculated, then this article proves that the constructed splitting authentication code is an optimal splitting authentication code; Finally, a concrete example of the constructed splitting authentication code is given, the successful impersonation attack probability and the successful substitution attack probability are calculated, the rationality of construction is verified by simulation, and verifies that it satisfies the optimality.
Splitting authentication codes are an important method to study authentication codes with arbitration. Splitting authentication codes have a higher utilization rate of encoding rules than non-splitting authentication codes. Splitting authentication codes are constructed through group divisible design in this article. Firstly, a theorem for constructing splitting authentication codes is given. The theorem uses Group Divisible Design (GDD) to construct a splitting-GDD, and then a splitting-Balanced Incomplete Block Design (BIBD) by splitting-GDD is constructed, and then a splitting authentication code is obtained; Secondly, the feasibility of constructing splitting authentication codes through GDD under the conditions given in this article is verified. Then a splitting design is given and a splitting authentication codes based on GDD is constructed; Thirdly, the number of sources, the number of encoding rules, the number of messages of the splitting authentication code, the impersonation attack probability and the substitution attack probability are calculated, then this article proves that the constructed splitting authentication code is an optimal splitting authentication code; Finally, a concrete example of the constructed splitting authentication code is given, the successful impersonation attack probability and the successful substitution attack probability are calculated, the rationality of construction is verified by simulation, and verifies that it satisfies the optimality.
2022, 44(2): 602-610.
doi: 10.11999/JEIT210152
Abstract:
The existing anomaly detection methods which require pre-learning and are sensitive to noise result in long detection time and high false positive rate. Based on the analysis of the existing anomaly detection cases, a new perspective is proposed from platform heterogeneity: programs are run on multiple heterogeneous platforms, normal programs are run on all platforms with the same result, while anomaly programs show heterogeneity on different platforms. So a lightweight program anomaly detection method for heterogeneous platforms is designed. System state data is collected. Feature engineering is used to construct a multidimensional vector with obvious representation of anomaly. The label code and max-min normalization are used to preprocess the data. The difference degree between the data is calculated and the threshold rule is used to compare, analyze and detect anomaly. Compared with the unsupervised feature clustering method, detection accuracy of the proposed method is improved by 13.12% with low false positive rate and short detection time.
The existing anomaly detection methods which require pre-learning and are sensitive to noise result in long detection time and high false positive rate. Based on the analysis of the existing anomaly detection cases, a new perspective is proposed from platform heterogeneity: programs are run on multiple heterogeneous platforms, normal programs are run on all platforms with the same result, while anomaly programs show heterogeneity on different platforms. So a lightweight program anomaly detection method for heterogeneous platforms is designed. System state data is collected. Feature engineering is used to construct a multidimensional vector with obvious representation of anomaly. The label code and max-min normalization are used to preprocess the data. The difference degree between the data is calculated and the threshold rule is used to compare, analyze and detect anomaly. Compared with the unsupervised feature clustering method, detection accuracy of the proposed method is improved by 13.12% with low false positive rate and short detection time.
2022, 44(2): 611-619.
doi: 10.11999/JEIT201050
Abstract:
As a typical application scenario of 5G uRLLC, the data transmission delay and reliability requirements of industrial applications are more and more stringent, and the convergent transmission of diversified data becomes an urgent problem to be solved. One of the important challenges is the efficient scheduling of wireless resources to ensure the coexistence of various data transmission without interfering with each other and stable operation of the system. In view of 5G uplink transmission in industrial transmission scenarios, a prediction-based resource allocation scheme is proposed, which uses Auto Regressive Moving Average (ARMA) model to predict the activation rates of the next transmission cycle based on the historical data. Then the resources are dynamically reserved for periodic and emergency data, so as to minimize the impact on periodic data transmission under the premise of meeting the emergency data transmission conditions. Experimental results show that, compared with the traditional resource allocation scheme, this scheme can effectively reduce the impact of emergency data transmission on periodic data and improve the utilization of spectrum resources.
As a typical application scenario of 5G uRLLC, the data transmission delay and reliability requirements of industrial applications are more and more stringent, and the convergent transmission of diversified data becomes an urgent problem to be solved. One of the important challenges is the efficient scheduling of wireless resources to ensure the coexistence of various data transmission without interfering with each other and stable operation of the system. In view of 5G uplink transmission in industrial transmission scenarios, a prediction-based resource allocation scheme is proposed, which uses Auto Regressive Moving Average (ARMA) model to predict the activation rates of the next transmission cycle based on the historical data. Then the resources are dynamically reserved for periodic and emergency data, so as to minimize the impact on periodic data transmission under the premise of meeting the emergency data transmission conditions. Experimental results show that, compared with the traditional resource allocation scheme, this scheme can effectively reduce the impact of emergency data transmission on periodic data and improve the utilization of spectrum resources.
2022, 44(2): 620-626.
doi: 10.11999/JEIT202019
Abstract:
In the millimeter wave Multiple Input Multiple Output (MIMO) system, in order to improve the spectral efficiency, an effective hybrid precoding double-layer alternating iterative algorithm is proposed. The alternation of the outer layer uses the decomposition method to decouple the transmitter and receiver, which reduces the number of solved variables in one calculation. The alternation of the inner layer is only used for the receiver or transmitter. The analog domain matrix is decomposed by columns and the digital domain matrix is decomposed by rows, which reduces the spectrum efficiency expression into a series of sub-problems. Considering the influence of single element in the analog domain matrix on the function and the limitation of amplitude 1, the column elements are optimized one by one, and each solution is constrained to make solution be always in the convergence domain. Experimental results show that the proposed alternative optimization solution can obtain better performance and low complexity.
In the millimeter wave Multiple Input Multiple Output (MIMO) system, in order to improve the spectral efficiency, an effective hybrid precoding double-layer alternating iterative algorithm is proposed. The alternation of the outer layer uses the decomposition method to decouple the transmitter and receiver, which reduces the number of solved variables in one calculation. The alternation of the inner layer is only used for the receiver or transmitter. The analog domain matrix is decomposed by columns and the digital domain matrix is decomposed by rows, which reduces the spectrum efficiency expression into a series of sub-problems. Considering the influence of single element in the analog domain matrix on the function and the limitation of amplitude 1, the column elements are optimized one by one, and each solution is constrained to make solution be always in the convergence domain. Experimental results show that the proposed alternative optimization solution can obtain better performance and low complexity.
2022, 44(2): 627-636.
doi: 10.11999/JEIT200698
Abstract:
In view of the limitation of the accuracy of the Angle of Arrival (AoA) and Time of Flight (ToF) estimation caused by limited array antennas and channel bandwidth in commodity Wi-Fi devices, and the maximum likelihood estimation algorithm tends sto converge to the local optimal value, and the way to estimate the parameters in turn will affect each other, a three-dimensional joint parameter estimation algorithm for AoA, ToF and Doppler Frequency Shift (DFS) based on Channel State Information (CSI) is proposed to improve parameter estimation accuracy and signal resolution. Firstly, a three-dimensional matrix containing AoA, ToF and DFS information is constructed and the dimensionality of the constructed three-dimensional matrix is reduced. Then, a three-dimensional spatial smoothing algorithm is designed to eliminate the influence of coherent signals on parameter estimation. Finally, the spatial spectrum estimation algorithm is carried out and the three parameters AoA, ToF and DFS are estimated simultaneously. Experimental results show that the accuracy and signal resolution of 3D joint parameter estimation method are higher than Space-Alternating Generalized Expectation maximization(SAGE) and two-dimensional parameter estimation methods.
In view of the limitation of the accuracy of the Angle of Arrival (AoA) and Time of Flight (ToF) estimation caused by limited array antennas and channel bandwidth in commodity Wi-Fi devices, and the maximum likelihood estimation algorithm tends sto converge to the local optimal value, and the way to estimate the parameters in turn will affect each other, a three-dimensional joint parameter estimation algorithm for AoA, ToF and Doppler Frequency Shift (DFS) based on Channel State Information (CSI) is proposed to improve parameter estimation accuracy and signal resolution. Firstly, a three-dimensional matrix containing AoA, ToF and DFS information is constructed and the dimensionality of the constructed three-dimensional matrix is reduced. Then, a three-dimensional spatial smoothing algorithm is designed to eliminate the influence of coherent signals on parameter estimation. Finally, the spatial spectrum estimation algorithm is carried out and the three parameters AoA, ToF and DFS are estimated simultaneously. Experimental results show that the accuracy and signal resolution of 3D joint parameter estimation method are higher than Space-Alternating Generalized Expectation maximization(SAGE) and two-dimensional parameter estimation methods.
2022, 44(2): 637-645.
doi: 10.11999/JEIT200840
Abstract:
Polar codes have perfect coding and decoding performance as a kind of error correction code, which have become a standard coding scheme for 5G short code control channel. While the length of polar codes is short, its performance is not good enough. A novel concatenating scheme, parity check codes concatenating polar codes, has improved the performance of the limited length of polar codes. However, its decoding algorithm has high complexity. In order to solve the problem, a Parity Check aided Partial Successive Cancellation List(PC-PSCL) algorithm based on parity-check-concatenated polar codes is proposed. In this algorithm, outer codes are constructed before encoding and the information bits with not enough reliability are selected through sub-channel error probability obtained by Gaussian Approximation (GA), which perform Successive Cancellation List (SCL) decoding with the help of parity check codes, while the remaining bits just perform Successive Cancellation (SC) decoding. The simulations in additive white Gaussian noise channel reveal that when the codes length is 512, the codes rate is 1/2, the frame error rate is\begin{document}$ {10}^{-3} $\end{document} ![]()
![]()
Polar codes have perfect coding and decoding performance as a kind of error correction code, which have become a standard coding scheme for 5G short code control channel. While the length of polar codes is short, its performance is not good enough. A novel concatenating scheme, parity check codes concatenating polar codes, has improved the performance of the limited length of polar codes. However, its decoding algorithm has high complexity. In order to solve the problem, a Parity Check aided Partial Successive Cancellation List(PC-PSCL) algorithm based on parity-check-concatenated polar codes is proposed. In this algorithm, outer codes are constructed before encoding and the information bits with not enough reliability are selected through sub-channel error probability obtained by Gaussian Approximation (GA), which perform Successive Cancellation List (SCL) decoding with the help of parity check codes, while the remaining bits just perform Successive Cancellation (SC) decoding. The simulations in additive white Gaussian noise channel reveal that when the codes length is 512, the codes rate is 1/2, the frame error rate is
2022, 44(2): 646-653.
doi: 10.11999/JEIT210561
Abstract:
In the Orthogonal Time-Frequency Space (OTFS) modulation system, it is difficult to estimate the Channel State Information (CSI) of physical path corresponding to fractional Doppler channel, and the computation is very complicated. To solve these problems, a channel estimation algorithm PRS-PMF (Pilot Resource Saving-Pulse Matched Filtering) for pulse matching filter which saves pilot resources is proposed. In the algorithm, the embedded auxiliary pilot is employed to obtain the equivalent channel estimation, then the CSI of each path is estimated through the cross-correlation matched filter. Compared with the traditional cross-correlation matched filter channel estimation algorithms, it can reduce the pilot resource occupy and the computational complexity. On this basis, the OTFS system is windowed to reduce the number of integer samples of the main lobe of the window response and reduce the side lobe level, which improves effectively the autocorrelation characteristics of the equivalent channel Doppler response function and thus reduces the interference of other symbols and noise on the estimated symbols.
In the Orthogonal Time-Frequency Space (OTFS) modulation system, it is difficult to estimate the Channel State Information (CSI) of physical path corresponding to fractional Doppler channel, and the computation is very complicated. To solve these problems, a channel estimation algorithm PRS-PMF (Pilot Resource Saving-Pulse Matched Filtering) for pulse matching filter which saves pilot resources is proposed. In the algorithm, the embedded auxiliary pilot is employed to obtain the equivalent channel estimation, then the CSI of each path is estimated through the cross-correlation matched filter. Compared with the traditional cross-correlation matched filter channel estimation algorithms, it can reduce the pilot resource occupy and the computational complexity. On this basis, the OTFS system is windowed to reduce the number of integer samples of the main lobe of the window response and reduce the side lobe level, which improves effectively the autocorrelation characteristics of the equivalent channel Doppler response function and thus reduces the interference of other symbols and noise on the estimated symbols.
Lawson-norm Constrained Generalized Lncosh Based Adaptive Algorithm for Sparse System Identification
2022, 44(2): 654-660.
doi: 10.11999/JEIT210057
Abstract:
A generalized Lawson-lncosh adaptive filtering algorithm for sparse system identification is proposed. The proposed algorithm is derived by constructing a new cost function consisted of Lawson-norm of system coefficient vector and lncosh function of instantaneous error. And the Lawson-norm constraint introduces a parameter p which can realize the dynamic adjustment of sparsity. The proposed algorithm can improve the convergence speed and reduce the steady-state error for sqarse system identification, where the Lncosh function of the error has the property of combating impulsive noise. Then, the range of the step-size and effect of parameter p on the proposed algorithm are analyzed. Computer simulation results show that the proposed algorithm has superior performance compared with other existing algorithms with Gaussian and colored input signals and the sparsity constraint for the proposed algorithm is controllable.
A generalized Lawson-lncosh adaptive filtering algorithm for sparse system identification is proposed. The proposed algorithm is derived by constructing a new cost function consisted of Lawson-norm of system coefficient vector and lncosh function of instantaneous error. And the Lawson-norm constraint introduces a parameter p which can realize the dynamic adjustment of sparsity. The proposed algorithm can improve the convergence speed and reduce the steady-state error for sqarse system identification, where the Lncosh function of the error has the property of combating impulsive noise. Then, the range of the step-size and effect of parameter p on the proposed algorithm are analyzed. Computer simulation results show that the proposed algorithm has superior performance compared with other existing algorithms with Gaussian and colored input signals and the sparsity constraint for the proposed algorithm is controllable.
2022, 44(2): 661-667.
doi: 10.11999/JEIT210073
Abstract:
Under\begin{document}$ \alpha $\end{document} ![]()
![]()
\begin{document}$ \alpha $\end{document} ![]()
![]()
Under
2022, 44(2): 668-676.
doi: 10.11999/JEIT210079
Abstract:
Considering the problem that the Low Peak-to-average power ratio Filter Bank MultiCarrier (LP-FBMC) structure based on Discrete Fourier Transform(DFT) spread spreading requires additional transmission Side Information (SI), an optimized structure of without SI based on local phase rotation of constellation symbol sequence is proposed. Considering that SI is generated by LP-FBMC choosing four signal transmission forms at the transmitter, combining the connection between the demodulated constellation symbols sequence and the original symbols sequence of these four forms, the local phase rotation of the constellation symbol sequence is adopted to avoid the transmission of SI. The local rotation method reduces the number of the phase rotation angles, which reduces the complexity, and the correct rate of phase estimation is improved at the receiver by expanding the decision range of phase. The simulation results show that the proposed structure not only maintains the same PAPR suppression performance and similar BER performance as the LP-FBMC structure, but also reduces the computational complexity by about 20% compared with the embedded SI structure.
Considering the problem that the Low Peak-to-average power ratio Filter Bank MultiCarrier (LP-FBMC) structure based on Discrete Fourier Transform(DFT) spread spreading requires additional transmission Side Information (SI), an optimized structure of without SI based on local phase rotation of constellation symbol sequence is proposed. Considering that SI is generated by LP-FBMC choosing four signal transmission forms at the transmitter, combining the connection between the demodulated constellation symbols sequence and the original symbols sequence of these four forms, the local phase rotation of the constellation symbol sequence is adopted to avoid the transmission of SI. The local rotation method reduces the number of the phase rotation angles, which reduces the complexity, and the correct rate of phase estimation is improved at the receiver by expanding the decision range of phase. The simulation results show that the proposed structure not only maintains the same PAPR suppression performance and similar BER performance as the LP-FBMC structure, but also reduces the computational complexity by about 20% compared with the embedded SI structure.
2022, 44(2): 677-685.
doi: 10.11999/JEIT210013
Abstract:
A pair of mismatched sequences is called an almost mismatched complementary pair if their periodic autocorrelation functions sum up to a same nonzero integer for all out-of-phase time shifts. In this paper, a new balanced quadriphase almost mismatched complementary pair is proposed, the theoretical bound of balanced quadriphase almost complementary pair with prime length is proved by Gray mapping, based on the cyclotomic classes of order 4, the optimal balanced quadriphase almost mismatched complementary pair with prime length satisfied theoretical bound is constructed. The existence of quadriphase complementary pairs are expanded and compensated the deficiency of most of the existing ones only have even length at present by investigate in this paper.
A pair of mismatched sequences is called an almost mismatched complementary pair if their periodic autocorrelation functions sum up to a same nonzero integer for all out-of-phase time shifts. In this paper, a new balanced quadriphase almost mismatched complementary pair is proposed, the theoretical bound of balanced quadriphase almost complementary pair with prime length is proved by Gray mapping, based on the cyclotomic classes of order 4, the optimal balanced quadriphase almost mismatched complementary pair with prime length satisfied theoretical bound is constructed. The existence of quadriphase complementary pairs are expanded and compensated the deficiency of most of the existing ones only have even length at present by investigate in this paper.
2022, 44(2): 686-693.
doi: 10.11999/JEIT210028
Abstract:
Finite Impulse Response(FIR) filter is the main component of multi-carrier modulation system in wireless communication research. Considering the problem of FIR filter performance degradation caused by the finite word length effect, an FIR filter lattice structure is proposed to optimize the filter coefficient error caused by quantization, that is, to reduce the coefficient sensitivity. The state space structure is used to express the corresponding improved lattice structure coefficients, and the coefficient sensitivity expression is derived and analyzed. The simulation results show that the sensitivity of the improved lattice structure coefficient is related to the sampling period. Compared with the traditional lattice structure, under the constraints of quantization word length and sampling period, the frequency response characteristic curve of the improved lattice structure is closer to the ideal frequency response characteristic curve, the coefficient sensitivity is smaller, and the ability of resisting finite word length effect is better.
Finite Impulse Response(FIR) filter is the main component of multi-carrier modulation system in wireless communication research. Considering the problem of FIR filter performance degradation caused by the finite word length effect, an FIR filter lattice structure is proposed to optimize the filter coefficient error caused by quantization, that is, to reduce the coefficient sensitivity. The state space structure is used to express the corresponding improved lattice structure coefficients, and the coefficient sensitivity expression is derived and analyzed. The simulation results show that the sensitivity of the improved lattice structure coefficient is related to the sampling period. Compared with the traditional lattice structure, under the constraints of quantization word length and sampling period, the frequency response characteristic curve of the improved lattice structure is closer to the ideal frequency response characteristic curve, the coefficient sensitivity is smaller, and the ability of resisting finite word length effect is better.
2022, 44(2): 694-701.
doi: 10.11999/JEIT210106
Abstract:
With the foundation of multi-carrier modulation based on Prolate Spheroidal Wave Functions (PSWFs) with signal grouping optimization, the idea of dual-mode index modulation is introduced, and proposes a method of Multi-Carrier index Modulation based on PSWFs with Dual-Mode (DM-MCM-PSWFs) is introduced. The remaining sub-carriers that are not activated to load the modulation symbols generated by the second constellation to transmit additional information bits are used in this method, resulting in the improvements of efficiency of spectrum resources in multi-carrier modulation based on PSWFs with signal grouping optimization, and it improves effectively the system spectral efficiency and error performance. Theoretical and simulation analysis indicate that compared with multi-carrier modulation based on PSWFs with signal grouping optimization, the proposed method can attain higher system spectral efficiency and better system error performance at the expense of acceptable sacrifice of system complexity. When the bit rate is 1×10-5, n=7, and k=3, the system spectral efficiency and error performance of the proposed method can increase by 9.15% and 2.4 dB respectively.
With the foundation of multi-carrier modulation based on Prolate Spheroidal Wave Functions (PSWFs) with signal grouping optimization, the idea of dual-mode index modulation is introduced, and proposes a method of Multi-Carrier index Modulation based on PSWFs with Dual-Mode (DM-MCM-PSWFs) is introduced. The remaining sub-carriers that are not activated to load the modulation symbols generated by the second constellation to transmit additional information bits are used in this method, resulting in the improvements of efficiency of spectrum resources in multi-carrier modulation based on PSWFs with signal grouping optimization, and it improves effectively the system spectral efficiency and error performance. Theoretical and simulation analysis indicate that compared with multi-carrier modulation based on PSWFs with signal grouping optimization, the proposed method can attain higher system spectral efficiency and better system error performance at the expense of acceptable sacrifice of system complexity. When the bit rate is 1×10-5, n=7, and k=3, the system spectral efficiency and error performance of the proposed method can increase by 9.15% and 2.4 dB respectively.
2022, 44(2): 702-709.
doi: 10.11999/JEIT210107
Abstract:
Age of Information (AoI) is a novel metric that describes the timeliness of data delivery for time-sensitive applications, which measures the freshness of the most recently received packet from the perspective of destination node. In the multi-channel wireless network scenario with limited channel resources, the constraints of channel resources and link conflicts should be considered in the link scheduling with respect to AoI. To address this issue, in this paper, a time slot based scheduling method for data transmission to minimize the average AoI in the network is proposed. In this method, the optimization problem of AoI is first formulated into a Lyapunov optimization problem. Then the multi-channel conflict problem is converted to find the maximum matching policy of bipartite graph, which is solved by Kuhn-Munkres (KM) algorithm. Thus, a scheduling policy under constraints is obtained. The simulation results demonstrate that the proposed method optimizes effectively the average AoI and improves the freshness of data in the network.
Age of Information (AoI) is a novel metric that describes the timeliness of data delivery for time-sensitive applications, which measures the freshness of the most recently received packet from the perspective of destination node. In the multi-channel wireless network scenario with limited channel resources, the constraints of channel resources and link conflicts should be considered in the link scheduling with respect to AoI. To address this issue, in this paper, a time slot based scheduling method for data transmission to minimize the average AoI in the network is proposed. In this method, the optimization problem of AoI is first formulated into a Lyapunov optimization problem. Then the multi-channel conflict problem is converted to find the maximum matching policy of bipartite graph, which is solved by Kuhn-Munkres (KM) algorithm. Thus, a scheduling policy under constraints is obtained. The simulation results demonstrate that the proposed method optimizes effectively the average AoI and improves the freshness of data in the network.
2022, 44(2): 710-717.
doi: 10.11999/JEIT210142
Abstract:
A low complexity Two-Dimensional Direction Of Arrival (2D-DOA) estimation algorithm using wideband Uniform Concentric Spherical Arrays (UCSA) is proposed in this paper. The output signals of the wideband UCSA are firstly converted to phase mode signals, which are then compensated by filters to achieve Frequency Invariant (FI) characteristic. The FI characteristic of the array can reduce the computational complexity of the wideband signal processing. To reduce further the complexity of 2D-DOA estimation, a reduced-dimension MUltiple SIgnal Classification (MUSIC) algorithm using FI-UCSA is proposed. The equivalent steering vector of the phase mode signal is decomposed into two matrices, which are related to azimuth and elevation angles, respectively. The 2D MUSIC is then simplified to One-Dimensional (1D) search, which optimizes DOA estimation in reduced dimension and substantially lower the computation complexity. Simulation results show that the computational complexity of the proposed algorithm is greatly reduced compared with the 2D MUSIC algorithm. In addition, the estimation accuracy and the resolution are slightly improved.
A low complexity Two-Dimensional Direction Of Arrival (2D-DOA) estimation algorithm using wideband Uniform Concentric Spherical Arrays (UCSA) is proposed in this paper. The output signals of the wideband UCSA are firstly converted to phase mode signals, which are then compensated by filters to achieve Frequency Invariant (FI) characteristic. The FI characteristic of the array can reduce the computational complexity of the wideband signal processing. To reduce further the complexity of 2D-DOA estimation, a reduced-dimension MUltiple SIgnal Classification (MUSIC) algorithm using FI-UCSA is proposed. The equivalent steering vector of the phase mode signal is decomposed into two matrices, which are related to azimuth and elevation angles, respectively. The 2D MUSIC is then simplified to One-Dimensional (1D) search, which optimizes DOA estimation in reduced dimension and substantially lower the computation complexity. Simulation results show that the computational complexity of the proposed algorithm is greatly reduced compared with the 2D MUSIC algorithm. In addition, the estimation accuracy and the resolution are slightly improved.
Low Probability of Intercept Radar Signal Detection Algorithm Based on Convolutional Neural Networks
2022, 44(2): 718-725.
doi: 10.11999/JEIT210132
Abstract:
In order to solve the problem of radar intercepting receiver’s unsatisfactory detection effect on Low Probability of Intercept (LPI) radar signal, a method of LPI radar signal detection based on Convolutional Neural Networks (CNN) is proposed, which defines signal and noise by effective signal pulse width in intercepted signal. The similarity of the convolution kernel and the matched filter in the structure can improve the detection accuracy of the signal under the low SNR.A large number of analog data sets based on four typical LPI radar signals (Linear Frequency Modulation signal (LFM), NonLinear Frequency Modulation signal (NLFM), Binary Phase Shift Keying signal (BPSK), COSTAS frequency coded signal) and white noise signals are used for CNN model training. At the same time, a small amount of measured signals (Linear Frequency Modulation signal (LFM), Binary Phase Shift Keying signal (BPSK)) are added as verification set for adaptation, so as to match better the detection model of measured signals. Finally, the experimental results show that the proposed algorithm has a good detection effect in the case of low SNR, and has the ability to generalize the LPI radar signals under various modulation modes and different SNR.
In order to solve the problem of radar intercepting receiver’s unsatisfactory detection effect on Low Probability of Intercept (LPI) radar signal, a method of LPI radar signal detection based on Convolutional Neural Networks (CNN) is proposed, which defines signal and noise by effective signal pulse width in intercepted signal. The similarity of the convolution kernel and the matched filter in the structure can improve the detection accuracy of the signal under the low SNR.A large number of analog data sets based on four typical LPI radar signals (Linear Frequency Modulation signal (LFM), NonLinear Frequency Modulation signal (NLFM), Binary Phase Shift Keying signal (BPSK), COSTAS frequency coded signal) and white noise signals are used for CNN model training. At the same time, a small amount of measured signals (Linear Frequency Modulation signal (LFM), Binary Phase Shift Keying signal (BPSK)) are added as verification set for adaptation, so as to match better the detection model of measured signals. Finally, the experimental results show that the proposed algorithm has a good detection effect in the case of low SNR, and has the ability to generalize the LPI radar signals under various modulation modes and different SNR.
2022, 44(2): 726-736.
doi: 10.11999/JEIT210078
Abstract:
The Sound Speed Profiles (SSPs) in sea water have obvious time evolution characteristics, and their prediction can be regarded as a nonlinear time series prediction. Recurrent Neural Networks (RNN), a type of deep neural network designed for sequence modeling, can capture nonlinear relationships flexibly. Attention Mechanism (AM), which selects the most critical information for the current task, can describe the nonlinear relationships in space and temporal dimensions. In this paper, RNN and AM are used to construct a multivariate time series prediction model to learn the historical SSPs and predict the time-varying full-sea SSPs in shallow sea environment. Experiments on real sound speed data show that the proposed method is effective and outperforms other methods, and provides a new idea for the combination of physical model and machine learning in underwater acoustics.
The Sound Speed Profiles (SSPs) in sea water have obvious time evolution characteristics, and their prediction can be regarded as a nonlinear time series prediction. Recurrent Neural Networks (RNN), a type of deep neural network designed for sequence modeling, can capture nonlinear relationships flexibly. Attention Mechanism (AM), which selects the most critical information for the current task, can describe the nonlinear relationships in space and temporal dimensions. In this paper, RNN and AM are used to construct a multivariate time series prediction model to learn the historical SSPs and predict the time-varying full-sea SSPs in shallow sea environment. Experiments on real sound speed data show that the proposed method is effective and outperforms other methods, and provides a new idea for the combination of physical model and machine learning in underwater acoustics.
2022, 44(2): 737-744.
doi: 10.11999/JEIT200912
Abstract:
Compared with the traditional high power microwave sources based on the electronic vacuum tube, the wide-band high power microwave sources based on the Gyromagnetic NonLinear Transmission Line (GNLTL) does not need the electron beam, the confining magnetic field and the vacuum system. It is a simple, compact and solid-state scheme. It has the advantages of adjustable frequency and repetitive operation. It can not only improve the energy utilization rate, but also break through the limitation of single operation of electromagnetic pulse projectile. In this paper, the RF pulse formation dynamics and sharpening characteristics of the gyromagnetic nonlinear transmission line are analyzed in theory. The visible two-dimensional model is composed for numerical analysis of the modulated pulse waveforms in time domain and frequency domain under a varied incident voltage or a varied axial biasing magnetic field. The simulation results show that when the incident voltage increases, the modulated peak voltage grows, the modulated depth rises and then declines to some constant value, the rise time of the modulated pulse reduces and then ceases, and the central frequency augments. When the axial biasing magnetic field increases, the modulated peak voltage and the modulated depth both rises and then declines, the rise time of the modulated pulse and the central frequency both reduces and then increases.
Compared with the traditional high power microwave sources based on the electronic vacuum tube, the wide-band high power microwave sources based on the Gyromagnetic NonLinear Transmission Line (GNLTL) does not need the electron beam, the confining magnetic field and the vacuum system. It is a simple, compact and solid-state scheme. It has the advantages of adjustable frequency and repetitive operation. It can not only improve the energy utilization rate, but also break through the limitation of single operation of electromagnetic pulse projectile. In this paper, the RF pulse formation dynamics and sharpening characteristics of the gyromagnetic nonlinear transmission line are analyzed in theory. The visible two-dimensional model is composed for numerical analysis of the modulated pulse waveforms in time domain and frequency domain under a varied incident voltage or a varied axial biasing magnetic field. The simulation results show that when the incident voltage increases, the modulated peak voltage grows, the modulated depth rises and then declines to some constant value, the rise time of the modulated pulse reduces and then ceases, and the central frequency augments. When the axial biasing magnetic field increases, the modulated peak voltage and the modulated depth both rises and then declines, the rise time of the modulated pulse and the central frequency both reduces and then increases.
2022, 44(2): 745-753.
doi: 10.11999/JEIT210029
Abstract:
Considering the problem that the similarity of magnetic anomaly signals is difficult to measure under low signal-to-noise ratio, a similarity measurement method OBF-EDR based on the combination of Orthogonal Basis Function (OBF) decomposition and Edit Distance on Real sequence (EDR) is proposed. This method obtains discrete basis function coefficients by decomposing the magnetic anomaly signals with orthogonal basis functions method. The signal-to-Noise Ratio (SNR) of discrete basis function coefficients is improved due to the uncorrelated characteristics of background noise and basis functions. EDR is used to measure the discrete coefficients of the basis function so as to measure indirectly the similarity of the magnetic anomaly signals. The simulation test shows that the OBF-EDR method can measure the similarity of magnetic anomaly signals at a lower SNR than the EDR algorithm.
Considering the problem that the similarity of magnetic anomaly signals is difficult to measure under low signal-to-noise ratio, a similarity measurement method OBF-EDR based on the combination of Orthogonal Basis Function (OBF) decomposition and Edit Distance on Real sequence (EDR) is proposed. This method obtains discrete basis function coefficients by decomposing the magnetic anomaly signals with orthogonal basis functions method. The signal-to-Noise Ratio (SNR) of discrete basis function coefficients is improved due to the uncorrelated characteristics of background noise and basis functions. EDR is used to measure the discrete coefficients of the basis function so as to measure indirectly the similarity of the magnetic anomaly signals. The simulation test shows that the OBF-EDR method can measure the similarity of magnetic anomaly signals at a lower SNR than the EDR algorithm.
2022, 44(2): 754-759.
doi: 10.11999/JEIT210037
Abstract:
By introducing the 5th-order term of power series expansion, the nonlinear distortion of the system under dual -frequency interference is analyzed, and the essential reason of dual-frequency insensitive effect caused by blocking interference is revealed, which is verified by experiments. Theoretical analysis and experimental results show that when the interference signal strength is weak and the nonlinear distortion of the system is low, the transfer function can be described by the power series expansion accurate to the 3rd-order term, and the test equipment is sensitive to the effective value of the dual -frequency interference field strength; with the increase of the interference signal, the nonlinear distortion of the system increases, and the 5th-order term in the power series expansion cannot be ignored, and the higher the blocking degree is, the more serious the dual-frequency insensitive effect is.
By introducing the 5th-order term of power series expansion, the nonlinear distortion of the system under dual -frequency interference is analyzed, and the essential reason of dual-frequency insensitive effect caused by blocking interference is revealed, which is verified by experiments. Theoretical analysis and experimental results show that when the interference signal strength is weak and the nonlinear distortion of the system is low, the transfer function can be described by the power series expansion accurate to the 3rd-order term, and the test equipment is sensitive to the effective value of the dual -frequency interference field strength; with the increase of the interference signal, the nonlinear distortion of the system increases, and the 5th-order term in the power series expansion cannot be ignored, and the higher the blocking degree is, the more serious the dual-frequency insensitive effect is.
2022, 44(2): 760-766.
doi: 10.11999/JEIT201034
Abstract:
The mean of spatial coverage number, the percentage of global coverage, and the mean of revisiting time are defined based on first-order statistics to evaluate spatiotemporal performance of spaceborne Global Navigation Satellites System (GNSS) reflectometry, in addition develops the GNSS and LEO orbits to simulate the influence of LEO orbit height, inclination and right ascension of ascending node on spatiotemporal performance. The results show that the mean of spatial coverage number, the percentage of global coverage, and the mean of revisiting time are respectively 1.6, 36.5, and 8 h with the orbit height of 1300 km, inclination of 98.7°, and the beam width of 40° so that the requirement of global observation could not be met. When the number of satellites in same orbital plane is 4 and the orbit height, inclination, and the beam width are respectively 1300 km, 98.7°, 40°, the spatiotemporal performance is higher than the 65% of single ASCAT. For the case of 8 satellites, spaceborne GNSS reflectometry could observe the area of 88.9% of the earth in 4.5 h.
The mean of spatial coverage number, the percentage of global coverage, and the mean of revisiting time are defined based on first-order statistics to evaluate spatiotemporal performance of spaceborne Global Navigation Satellites System (GNSS) reflectometry, in addition develops the GNSS and LEO orbits to simulate the influence of LEO orbit height, inclination and right ascension of ascending node on spatiotemporal performance. The results show that the mean of spatial coverage number, the percentage of global coverage, and the mean of revisiting time are respectively 1.6, 36.5, and 8 h with the orbit height of 1300 km, inclination of 98.7°, and the beam width of 40° so that the requirement of global observation could not be met. When the number of satellites in same orbital plane is 4 and the orbit height, inclination, and the beam width are respectively 1300 km, 98.7°, 40°, the spatiotemporal performance is higher than the 65% of single ASCAT. For the case of 8 satellites, spaceborne GNSS reflectometry could observe the area of 88.9% of the earth in 4.5 h.
2022, 44(2): 775-780.
doi: 10.11999/JEIT210006
Abstract:
When updating body’s orientation by using Micro Electro-Mechanical System Inertial Measurement Unit (MEMS-IMU), a precision turntable is needed and measurement errors are difficult to be estimated accurately, which cause high cost and large error. Based on the problem, both body’s attitude and gyroscope’s bias are estimated in field by a Kalman filter, an estimation method of measurement errors is proposed for improving orientation’s accuracy. Measurement errors for the optimal orientation’s accuracy are obtained by deducing the measurement errors’ mathematical model, and analyzing the relationship between orientation’s accuracy and the variations of measurement errors. By using a MEMS-IMU, the body’s orientation is measured when it keeps movement at a constant speed and in arbitrary trajectory for 5 minutes. Experimental results show the orientation after optimization is consistent with the referenced orientation, and pitch, roll and yaw during the last quiescent period deviate from the reference values only by 0.008°, 0.006° and 0.6° respectively.
When updating body’s orientation by using Micro Electro-Mechanical System Inertial Measurement Unit (MEMS-IMU), a precision turntable is needed and measurement errors are difficult to be estimated accurately, which cause high cost and large error. Based on the problem, both body’s attitude and gyroscope’s bias are estimated in field by a Kalman filter, an estimation method of measurement errors is proposed for improving orientation’s accuracy. Measurement errors for the optimal orientation’s accuracy are obtained by deducing the measurement errors’ mathematical model, and analyzing the relationship between orientation’s accuracy and the variations of measurement errors. By using a MEMS-IMU, the body’s orientation is measured when it keeps movement at a constant speed and in arbitrary trajectory for 5 minutes. Experimental results show the orientation after optimization is consistent with the referenced orientation, and pitch, roll and yaw during the last quiescent period deviate from the reference values only by 0.008°, 0.006° and 0.6° respectively.
