Email alert
2021 Vol. 43, No. 9
Display Method:
2021, 43(9): 2439-2448.
doi: 10.11999/JEIT210581
Abstract:
The power-analysis countermeasure for Elliptic Curve Cryptographic (ECC) chips endures large area, power consumption and performance degradation. In this paper, the difference in the probability distribution of the intermediate data Hamming distance is analyzed when the key guess is correct and incorrect in the point multiplication of ECC. A power compensation method based on dynamic Hamming distance control is proposed, which uses the simulated annealing algorithm offline to find the optimal mapping matrix. Finally, a mapping compensation model of equal probability on the elliptic curve cryptographic hardware is formed, which greatly reduces the correlation between intermediate data and power consumption. At the same time, a low-cost synchronous power compensation circuit is designed in the guidance of this model. Under the CMOS 40 nm process, the area of protected ECC128 is only increased by 22.8%. Experiments and tests are carried out on the Sakura-G board. The power overhead is 18.8%, and the number of minimum leakage traces is greater than 104, which is increased by 312 times. This countermeasure is the same as randomization with low cost and no impact on the throughput rate, which is suitable for high-speed or resource-constrained ECC circuits.
The power-analysis countermeasure for Elliptic Curve Cryptographic (ECC) chips endures large area, power consumption and performance degradation. In this paper, the difference in the probability distribution of the intermediate data Hamming distance is analyzed when the key guess is correct and incorrect in the point multiplication of ECC. A power compensation method based on dynamic Hamming distance control is proposed, which uses the simulated annealing algorithm offline to find the optimal mapping matrix. Finally, a mapping compensation model of equal probability on the elliptic curve cryptographic hardware is formed, which greatly reduces the correlation between intermediate data and power consumption. At the same time, a low-cost synchronous power compensation circuit is designed in the guidance of this model. Under the CMOS 40 nm process, the area of protected ECC128 is only increased by 22.8%. Experiments and tests are carried out on the Sakura-G board. The power overhead is 18.8%, and the number of minimum leakage traces is greater than 104, which is increased by 312 times. This countermeasure is the same as randomization with low cost and no impact on the throughput rate, which is suitable for high-speed or resource-constrained ECC circuits.
2021, 43(9): 2449-2457.
doi: 10.11999/JEIT210566
Abstract:
To investigate the effect of ElectroMagnetic Fault Injection (EMFI) attacks on the security of Dynamic Random Access Memory (DRAM), DRAMs are applied to an EMFI platform and are attacked with same attack settings firstly in this paper. Firstly, the attack results are collected and the faults are classified. Secondly, the mechanism of the faults occurred in the experiments are analyzed based on the fundamental structure of DRAM. Finally, the threats of the faults occurred in DRAM to the security of computer system are analyzed. In the experiments, multiple transient faults and multiple persistent faults are found. According to the experiments and analysis, it is found that the EMFI can inject persistent faults into the specified addresses of DRAM with low spatio-temporal resolution. In addition, persistent faults are successfully injected into S-box of AES-128 that stored in DRAM in this paper, and the key of AES is recovered by exploiting the faults. The experiment of key cracking indicates that EMFI attacks on DRAM pose serious security threats to computer systems, and the experiment indicates that researching on the security of DRAM is of great significance to hardware security.
To investigate the effect of ElectroMagnetic Fault Injection (EMFI) attacks on the security of Dynamic Random Access Memory (DRAM), DRAMs are applied to an EMFI platform and are attacked with same attack settings firstly in this paper. Firstly, the attack results are collected and the faults are classified. Secondly, the mechanism of the faults occurred in the experiments are analyzed based on the fundamental structure of DRAM. Finally, the threats of the faults occurred in DRAM to the security of computer system are analyzed. In the experiments, multiple transient faults and multiple persistent faults are found. According to the experiments and analysis, it is found that the EMFI can inject persistent faults into the specified addresses of DRAM with low spatio-temporal resolution. In addition, persistent faults are successfully injected into S-box of AES-128 that stored in DRAM in this paper, and the key of AES is recovered by exploiting the faults. The experiment of key cracking indicates that EMFI attacks on DRAM pose serious security threats to computer systems, and the experiment indicates that researching on the security of DRAM is of great significance to hardware security.
2021, 43(9): 2458-2465.
doi: 10.11999/JEIT210631
Abstract:
Most of the reported hardware obfuscations are single-level ones focusing on physical level, logical level or behavior level, in which the lack of synergy among different levels commonly results in limited security performance. Based on study of the relationships among circuit layout, logic and states transition, a multi-level co-obfuscation scheme is proposed to protect hardware IP cores. In bottom-up collaborative confusion design, dummy vias are introduced into camouflage gates layout to perform physical-logic obfuscation, and via-PUF (Physical Unclonable Fuction) are utilized in state transition control to realize physical-behavior obfuscation. Then, in top-down collaborative obfuscation design, logic locks are used to perform behavior-logic obfuscation, and parallel-branch obfuscation wire technique is designed to complete the behavior-physical confusion. Finally, a substitution algorithm of the obfuscation gates into the circuit’s netlist is proposed, and the three-level cooperative obfuscation is realized to achieve IP core security protection. ISCAS-89 Benchmarks and a typical cryptogram algorithm are used to verify the correctness and efficiency of the proposed IP core protection scheme. The test results show that under TSMC 65nm process, the average area cost percentage of the proposed co-obfuscation in large-scale circuits is 11.7%, the average power consumption accounts for 5.1%, The difference of register toggle between correct and wrong keys is less than 10%, and the proposed scheme can effectively resist violence attack, reverse engineering, boolean SATisfiability (SAT) attack.
Most of the reported hardware obfuscations are single-level ones focusing on physical level, logical level or behavior level, in which the lack of synergy among different levels commonly results in limited security performance. Based on study of the relationships among circuit layout, logic and states transition, a multi-level co-obfuscation scheme is proposed to protect hardware IP cores. In bottom-up collaborative confusion design, dummy vias are introduced into camouflage gates layout to perform physical-logic obfuscation, and via-PUF (Physical Unclonable Fuction) are utilized in state transition control to realize physical-behavior obfuscation. Then, in top-down collaborative obfuscation design, logic locks are used to perform behavior-logic obfuscation, and parallel-branch obfuscation wire technique is designed to complete the behavior-physical confusion. Finally, a substitution algorithm of the obfuscation gates into the circuit’s netlist is proposed, and the three-level cooperative obfuscation is realized to achieve IP core security protection. ISCAS-89 Benchmarks and a typical cryptogram algorithm are used to verify the correctness and efficiency of the proposed IP core protection scheme. The test results show that under TSMC 65nm process, the average area cost percentage of the proposed co-obfuscation in large-scale circuits is 11.7%, the average power consumption accounts for 5.1%, The difference of register toggle between correct and wrong keys is less than 10%, and the proposed scheme can effectively resist violence attack, reverse engineering, boolean SATisfiability (SAT) attack.
2021, 43(9): 2466-2473.
doi: 10.11999/JEIT210577
Abstract:
With illegal hardware reverse engineering attacks, the Integrated Circuit (IC) design suffers from the key Intellectual Property (IP)/algorithm piracy and hardware Trojan insertion. An IC camouflaging method, LPerturb, is proposed in this paper by local circuit logic perturbation for IP Protection. The circuit is partitioned into some Maximum Fanout-Free Cones (MFFCs), namely multiple functionally independent sub-circuits to be camouflaged, for output logic perturbation locally. A logic cell is selected in the MFFC sub-circuit. The cell is replaced to perturb the logic functionality of the MFFC minimally. A multi-logic camouflaged block is used to protect and restore the perturbed logic secret. Experimental results show that LPerturb can produce the camouflaged circuits steadily, which has good output corruptibility and effectively resists SAT based attack. The overhead in area and timing is also in low level.
With illegal hardware reverse engineering attacks, the Integrated Circuit (IC) design suffers from the key Intellectual Property (IP)/algorithm piracy and hardware Trojan insertion. An IC camouflaging method, LPerturb, is proposed in this paper by local circuit logic perturbation for IP Protection. The circuit is partitioned into some Maximum Fanout-Free Cones (MFFCs), namely multiple functionally independent sub-circuits to be camouflaged, for output logic perturbation locally. A logic cell is selected in the MFFC sub-circuit. The cell is replaced to perturb the logic functionality of the MFFC minimally. A multi-logic camouflaged block is used to protect and restore the perturbed logic secret. Experimental results show that LPerturb can produce the camouflaged circuits steadily, which has good output corruptibility and effectively resists SAT based attack. The overhead in area and timing is also in low level.
2021, 43(9): 2474-2481.
doi: 10.11999/JEIT210726
Abstract:
Physical Unclonable Function (PUF) has extremely important application prospects to the field of information security, however, there are also shortcomings in its own security from machine learning attacks and other aspects. By studying PUF circuits and cryptographic algorithm, a method based on sequence cipher of strong-PUF is proposed to resist machine learning attacks. Firstly, the random key is generated by constructing a rolling key generator, which is obfuscated with the input challenge; Then the obfuscated challenge is applied to the strong-PUF through a series-parallel conversion circuit to generate the output response; Finally, Python software simulation and FPGA hardware implementation are used to analyze the safety and statistical properties. The experimental results show that the attack prediction rates based on logistic regression, artificial neural network and support vector machine are close to the ideal value of 50% when the CRPs used for modeling are up to 106 groups. In addition, this method has high versatile, low hardware overhead and does not affect the randomness, uniqueness and reliability of PUF.
Physical Unclonable Function (PUF) has extremely important application prospects to the field of information security, however, there are also shortcomings in its own security from machine learning attacks and other aspects. By studying PUF circuits and cryptographic algorithm, a method based on sequence cipher of strong-PUF is proposed to resist machine learning attacks. Firstly, the random key is generated by constructing a rolling key generator, which is obfuscated with the input challenge; Then the obfuscated challenge is applied to the strong-PUF through a series-parallel conversion circuit to generate the output response; Finally, Python software simulation and FPGA hardware implementation are used to analyze the safety and statistical properties. The experimental results show that the attack prediction rates based on logistic regression, artificial neural network and support vector machine are close to the ideal value of 50% when the CRPs used for modeling are up to 106 groups. In addition, this method has high versatile, low hardware overhead and does not affect the randomness, uniqueness and reliability of PUF.
2021, 43(9): 2482-2488.
doi: 10.11999/JEIT210624
Abstract:
System-on-Chip (SoC) designers typically use third Party Intellectual Property(3PIP) cores to implement target functions. As these 3PIP cores are not trusted, the underlying SoC suffers from the threat of Hardware Trojans(HTs). As a subset of design-for-trust techniques, the diversified redundancy is promising in establishing trustworthy computings of SoCs. However, It is shown that the diversified redundancy can be defeated by HTs that explores triggering patterns. Therefore, an adapted diversified redundancy technique is proposed to defend against such kind of attacks.
System-on-Chip (SoC) designers typically use third Party Intellectual Property(3PIP) cores to implement target functions. As these 3PIP cores are not trusted, the underlying SoC suffers from the threat of Hardware Trojans(HTs). As a subset of design-for-trust techniques, the diversified redundancy is promising in establishing trustworthy computings of SoCs. However, It is shown that the diversified redundancy can be defeated by HTs that explores triggering patterns. Therefore, an adapted diversified redundancy technique is proposed to defend against such kind of attacks.
2021, 43(9): 2489-2497.
doi: 10.11999/JEIT210588
Abstract:
The operation characteristics of the tower field is analyzed, a nonlinear transformation realization method based on the tower domain is proposed. A random mask schedule for the inversion operation is designed, and cyclic shift is used in the randomization of mask, forming cyclic shift random mask scheme based on the tower domain, realizing the randomized hiding of all intermediate values and improving the ability of the algorithm to resist power attacks. The method proposed is verified on the Advanced Encryption Standard (AES) algorithm with the use of T-test and correlation analysis to evaluate the security of the masking scheme. There is no obvious information leakage points in the schedule, proving the ability to effectively resist correlation attacks. In addition, compared with the mask schedule in existing reference, the mask schedule proposed in this paper has less resource overhead and better generality.
The operation characteristics of the tower field is analyzed, a nonlinear transformation realization method based on the tower domain is proposed. A random mask schedule for the inversion operation is designed, and cyclic shift is used in the randomization of mask, forming cyclic shift random mask scheme based on the tower domain, realizing the randomized hiding of all intermediate values and improving the ability of the algorithm to resist power attacks. The method proposed is verified on the Advanced Encryption Standard (AES) algorithm with the use of T-test and correlation analysis to evaluate the security of the masking scheme. There is no obvious information leakage points in the schedule, proving the ability to effectively resist correlation attacks. In addition, compared with the mask schedule in existing reference, the mask schedule proposed in this paper has less resource overhead and better generality.
2021, 43(9): 2498-2507.
doi: 10.11999/JEIT210614
Abstract:
In order to evaluate the security of Physical Unclonable Function (PUF), it is necessary to put forward corresponding attack methods for different PUF structures. By studying the structure and working mechanism of Flip-Flop based Arbiter Physical Unclonable Function (FF-APUF), an effective attack method against FF-APUF is proposed based on Artificial Neural Network (ANN) in this paper. Firstly, according to the circuit structure, the delay model of FF-APUF is established by using multidimensional array. Secondly, all binary challenge bits are divided by two adjacent bits which are converted to a decimal, and then the challenges are expressed as a row vector to extract the feature vector. Finally, based on the extracted feature vectors, the attack model is constructed by ANN, and the optimal parameters are obtained by back propagation algorithm. The experimental results show that the prediction accuracy of the proposed method is higher than other three common machine learning methods under the same conditions. The attack advantage is more obvious, especially when the number of Challenge Response Pairs (CRP) is less and the bit number of challenges is large. For example, when the number of challenge bit is 128, and the number of CRPs is 100 and 500, the average attack prediction accuracy increased by 36.0% and 16.1% respectively. In addition, the proposed method has good robustness and scalability, and the maximum difference of attack prediction rate and reliability is only 0.32% under different noise.
In order to evaluate the security of Physical Unclonable Function (PUF), it is necessary to put forward corresponding attack methods for different PUF structures. By studying the structure and working mechanism of Flip-Flop based Arbiter Physical Unclonable Function (FF-APUF), an effective attack method against FF-APUF is proposed based on Artificial Neural Network (ANN) in this paper. Firstly, according to the circuit structure, the delay model of FF-APUF is established by using multidimensional array. Secondly, all binary challenge bits are divided by two adjacent bits which are converted to a decimal, and then the challenges are expressed as a row vector to extract the feature vector. Finally, based on the extracted feature vectors, the attack model is constructed by ANN, and the optimal parameters are obtained by back propagation algorithm. The experimental results show that the prediction accuracy of the proposed method is higher than other three common machine learning methods under the same conditions. The attack advantage is more obvious, especially when the number of Challenge Response Pairs (CRP) is less and the bit number of challenges is large. For example, when the number of challenge bit is 128, and the number of CRPs is 100 and 500, the average attack prediction accuracy increased by 36.0% and 16.1% respectively. In addition, the proposed method has good robustness and scalability, and the maximum difference of attack prediction rate and reliability is only 0.32% under different noise.
2021, 43(9): 2508-2517.
doi: 10.11999/JEIT200379
Abstract:
As the feature size of integrated circuits continues to scale down, under the harsh radiation environment, the probability of single event triple node upsets in nano-scale CMOS integrated circuits is increasing, seriously affecting reliability. In order to realize the resilient of single-event triple-node-upsets, a Low-Cost Triple-Node-Upset-Resilient Latch (LC-TNURL) is proposed. The latch is composed of seven C-elements and seven clock-gating C-elements, and has a symmetrical ring-shaped cross-interlock structure. Using the interceptive characteristics of the C-elements and the cross-interlock connection mode, after any three internal nodes are flipped, the transient pulse propagates inside the latch. After the C-elements is blocked in multiple stages, it will disappear step by step to ensure the LC-TNURL latch can self-recover to the correct logic state. Detailed HSPICE simulation shows that the power consumption of the LC-TNURL latch is reduced by an average of 31.9%, the delay is reduced by an average of 87.8%, the power-delay product is reduced by an average of 92.3% and the area overhead is increased by an average of 15.4% compared to other triple-node-upsets hardened latches (TNU-Latch, LCTNUT, TNUTL, TNURL). The LC-TNURL latch proposed in this paper is the least sensitive to PVT fluctuations and has high reliability compared with reference latches.
As the feature size of integrated circuits continues to scale down, under the harsh radiation environment, the probability of single event triple node upsets in nano-scale CMOS integrated circuits is increasing, seriously affecting reliability. In order to realize the resilient of single-event triple-node-upsets, a Low-Cost Triple-Node-Upset-Resilient Latch (LC-TNURL) is proposed. The latch is composed of seven C-elements and seven clock-gating C-elements, and has a symmetrical ring-shaped cross-interlock structure. Using the interceptive characteristics of the C-elements and the cross-interlock connection mode, after any three internal nodes are flipped, the transient pulse propagates inside the latch. After the C-elements is blocked in multiple stages, it will disappear step by step to ensure the LC-TNURL latch can self-recover to the correct logic state. Detailed HSPICE simulation shows that the power consumption of the LC-TNURL latch is reduced by an average of 31.9%, the delay is reduced by an average of 87.8%, the power-delay product is reduced by an average of 92.3% and the area overhead is increased by an average of 15.4% compared to other triple-node-upsets hardened latches (TNU-Latch, LCTNUT, TNUTL, TNURL). The LC-TNURL latch proposed in this paper is the least sensitive to PVT fluctuations and has high reliability compared with reference latches.
2021, 43(9): 2518-2525.
doi: 10.11999/JEIT200807
Abstract:
In the fields requiring precise time and frequency measurement and control, high-precision, dead-time free time interval and frequency measurements are highly demanded. TDC (Time to Digital Converter) is commonly adopted in time and frequency measurement. In this paper, a self-developed multi-module time-interval measurement system is built based on the time digital conversion chip TDC-GP21 from ACAM company and the Cyclone IV FPGA chip EP4CE6E22C8N from Altera Company. For each time-interval measurement module, a time-interval measurement resolution as small as 13 ps is achieved. By further duplex operating two such time-interval measurement modules, a dead-time free frequency measurement is realized, the time and frequency measurement system involves three groups, each of which has 3 TDC chips. By averaging the measurement results of three TDC chips in each group, a frequency instability reaches\begin{document}$ 1.1\times {10}^{-11} $\end{document} ![]()
![]()
\begin{document}$ 5.6\times {10}^{-15} $\end{document} ![]()
![]()
In the fields requiring precise time and frequency measurement and control, high-precision, dead-time free time interval and frequency measurements are highly demanded. TDC (Time to Digital Converter) is commonly adopted in time and frequency measurement. In this paper, a self-developed multi-module time-interval measurement system is built based on the time digital conversion chip TDC-GP21 from ACAM company and the Cyclone IV FPGA chip EP4CE6E22C8N from Altera Company. For each time-interval measurement module, a time-interval measurement resolution as small as 13 ps is achieved. By further duplex operating two such time-interval measurement modules, a dead-time free frequency measurement is realized, the time and frequency measurement system involves three groups, each of which has 3 TDC chips. By averaging the measurement results of three TDC chips in each group, a frequency instability reaches
2021, 43(9): 2526-2533.
doi: 10.11999/JEIT200372
Abstract:
The allocation of computing resources and register resources is an important issue for automatic parallel mapping of reconfigurable processors. In this paper, for the resource allocation of reconfigurable block cipher instruction set processors, an operator scheduling parameter model and processor resource parameter model are established, and the constraint relationship between the parallel scheduling of block ciphers and resource consumption is studied; Consequently, an automatic mapping algorithm based on greedy thinking, list scheduling and linear scanning is proposed to realize the block cipher atuomatic mapping on the reconfigurable block cipher instruction set processor. The experiment verifies the effect of the algorithm’s parallel scheduling under different resource constraints, and the contrast of AES-128 algorithm’s mapping effect is made to verify the progress of the algorithm, which obtains certain significance for the parallel computing research of block ciphers in reconfigurable processing.
The allocation of computing resources and register resources is an important issue for automatic parallel mapping of reconfigurable processors. In this paper, for the resource allocation of reconfigurable block cipher instruction set processors, an operator scheduling parameter model and processor resource parameter model are established, and the constraint relationship between the parallel scheduling of block ciphers and resource consumption is studied; Consequently, an automatic mapping algorithm based on greedy thinking, list scheduling and linear scanning is proposed to realize the block cipher atuomatic mapping on the reconfigurable block cipher instruction set processor. The experiment verifies the effect of the algorithm’s parallel scheduling under different resource constraints, and the contrast of AES-128 algorithm’s mapping effect is made to verify the progress of the algorithm, which obtains certain significance for the parallel computing research of block ciphers in reconfigurable processing.
2021, 43(9): 2534-2541.
doi: 10.11999/JEIT200805
Abstract:
In cloud computing applications, it is very important to ensure the confidentiality and unforgeability of messages, while judging the equivalence of different ciphertexts to plaintexts. The signcryption scheme with equality test can achieve the above security goals. Based on the certificateless public key cryptography environment, a Certificateless SignCryption scheme with Equality Test (CLSCET) is designed. Firstly, the framework and security model of the certificateless signcryption with equality test scheme are proposed, moreover two types of adversaries with different attack capabilities and three types of security targets are defined. Secondly, a specific certificateless signcryption with equality test scheme is constructed, and the correctness of the scheme is analyzed. Finally, based on the random oracle model, it is proved that the scheme satisfies the security properties of One-Way against Chosen Ciphertext Attack(OW-CCA), INDistinguishability against adaptive Chosen Ciphertext Attack(IND-CCA2) and Existential UnForgeability against adaptive Chosen Message Attack(EUF-CMA). Compared with the existing approximate schemes, the scheme satisfies the confidentiality of IND-CCA2, the unforgeability of EUF-CMA and the one-way ciphertext of OW-CCA.
In cloud computing applications, it is very important to ensure the confidentiality and unforgeability of messages, while judging the equivalence of different ciphertexts to plaintexts. The signcryption scheme with equality test can achieve the above security goals. Based on the certificateless public key cryptography environment, a Certificateless SignCryption scheme with Equality Test (CLSCET) is designed. Firstly, the framework and security model of the certificateless signcryption with equality test scheme are proposed, moreover two types of adversaries with different attack capabilities and three types of security targets are defined. Secondly, a specific certificateless signcryption with equality test scheme is constructed, and the correctness of the scheme is analyzed. Finally, based on the random oracle model, it is proved that the scheme satisfies the security properties of One-Way against Chosen Ciphertext Attack(OW-CCA), INDistinguishability against adaptive Chosen Ciphertext Attack(IND-CCA2) and Existential UnForgeability against adaptive Chosen Message Attack(EUF-CMA). Compared with the existing approximate schemes, the scheme satisfies the confidentiality of IND-CCA2, the unforgeability of EUF-CMA and the one-way ciphertext of OW-CCA.
2021, 43(9): 2542-2551.
doi: 10.11999/JEIT200554
Abstract:
This paper proposes an adaptive image dehazing algorithm, which fully considers the image features in different complex scenes and establishes an adaptive mechanism of the algorithm. The mechanism includes adaptive adjustments to whether the image is foggy, whether it is a sky area, or filter size, etc., which solves the bad effect that the traditional algorithm may cause when dehazing the depth mutation region. This article also implements FPGA acceleration for the adaptive image dehazing algorithm. Experimental results show that the algorithm can meet the real-time requirements of 1080P@60Hz video dehazing on XC7K325T FPGA video processing platform. For most light fog or heavy fog scenes, the image color of this algorithm is naturally free of oversaturation after dehazing. The average global contrast and saturation enhancement ratio are 0.309 and 0.994, which has obvious advantages compared with other dehazing algorithms in the field.
This paper proposes an adaptive image dehazing algorithm, which fully considers the image features in different complex scenes and establishes an adaptive mechanism of the algorithm. The mechanism includes adaptive adjustments to whether the image is foggy, whether it is a sky area, or filter size, etc., which solves the bad effect that the traditional algorithm may cause when dehazing the depth mutation region. This article also implements FPGA acceleration for the adaptive image dehazing algorithm. Experimental results show that the algorithm can meet the real-time requirements of 1080P@60Hz video dehazing on XC7K325T FPGA video processing platform. For most light fog or heavy fog scenes, the image color of this algorithm is naturally free of oversaturation after dehazing. The average global contrast and saturation enhancement ratio are 0.309 and 0.994, which has obvious advantages compared with other dehazing algorithms in the field.
2021, 43(9): 2552-2560.
doi: 10.11999/JEIT200859
Abstract:
In view of the complexity and low accuracy of feature extraction of amputees’ movement gestures, a feature extraction method based on gray model is proposed in this paper. Firstly, the pre-processed surface ElectroMyoGraphy (sEMG) and acceleration signals are intercepted by sliding window. Then, the mean value of the surface EMG signal, the driving coefficient of the gray model and the absolute mean value of the acceleration signal are extracted as features to form a feature vector. Finally, the features of the signal intercepted by sliding window are identified continuously. The proposed method is verified using NinaPro (Non Invasive Adaptive Prosthetics) public dataset, experimental results show that the proposed algorithm can effectively extract the characteristics of the electromyography and acceleration signals. An average accuracy of 91.14% is reached for 17 action gestures of 9 amputation subjects. The proposed approach provides a new way for the control algorithm of bionic limbs based human-computer interaction.
In view of the complexity and low accuracy of feature extraction of amputees’ movement gestures, a feature extraction method based on gray model is proposed in this paper. Firstly, the pre-processed surface ElectroMyoGraphy (sEMG) and acceleration signals are intercepted by sliding window. Then, the mean value of the surface EMG signal, the driving coefficient of the gray model and the absolute mean value of the acceleration signal are extracted as features to form a feature vector. Finally, the features of the signal intercepted by sliding window are identified continuously. The proposed method is verified using NinaPro (Non Invasive Adaptive Prosthetics) public dataset, experimental results show that the proposed algorithm can effectively extract the characteristics of the electromyography and acceleration signals. An average accuracy of 91.14% is reached for 17 action gestures of 9 amputation subjects. The proposed approach provides a new way for the control algorithm of bionic limbs based human-computer interaction.
2021, 43(9): 2561-2568.
doi: 10.11999/JEIT200480
Abstract:
Acute inferior myocardial infarction is a kind of heart disease with rapid progression and high mortality. In order to improve the diagnosis efficiency for inferior myocardial infarction, a novel algorithm for automatic detection of inferior myocardial infarction based on Bi-directional Long Short-Term Memory (BiLSTM) network of morphological feature extraction is proposed. Based on the clinical ECG signals of the cardiology center, noise is reduced and every heartbeat is segmented. According to the cardiology clinical guidelines and signal analysis, 12 lead waveform distance features and single lead waveform amplitude features are extracted. Additionally, the neural network structure of Long Short-Term Memory (LSTM) and BiLSTM are built from to the extracted features. It is cross-validated by Physikalisch-Technische Bundesanstalt (PTB) public database and chest pain center database, the accuracy reaches 99.72%, the precision and sensitivity reach 99.53% and 100%. At the same time, the F1-Score reaches 99.76. Furthermore, experimental results demonstrated that the accuracy of the novel algorithm is still 1% higher than that of other existing algorithms after adding the chest pain center database.
Acute inferior myocardial infarction is a kind of heart disease with rapid progression and high mortality. In order to improve the diagnosis efficiency for inferior myocardial infarction, a novel algorithm for automatic detection of inferior myocardial infarction based on Bi-directional Long Short-Term Memory (BiLSTM) network of morphological feature extraction is proposed. Based on the clinical ECG signals of the cardiology center, noise is reduced and every heartbeat is segmented. According to the cardiology clinical guidelines and signal analysis, 12 lead waveform distance features and single lead waveform amplitude features are extracted. Additionally, the neural network structure of Long Short-Term Memory (LSTM) and BiLSTM are built from to the extracted features. It is cross-validated by Physikalisch-Technische Bundesanstalt (PTB) public database and chest pain center database, the accuracy reaches 99.72%, the precision and sensitivity reach 99.53% and 100%. At the same time, the F1-Score reaches 99.76. Furthermore, experimental results demonstrated that the accuracy of the novel algorithm is still 1% higher than that of other existing algorithms after adding the chest pain center database.
2021, 43(9): 2569-2576.
doi: 10.11999/JEIT200687
Abstract:
Object tracking is one of the important research fields in computer vision. However, most tracking algorithm can not effectively learn the features suitable for tracking scene, which limits the performance improvement of tracking algorithm. To overcome this problem, this paper proposes a target tracking algorithm based on CNN Spatial and Channel Attention Mechanisms (CNNSCAM). The method consists of an off-line training apparent model and an adaptive updating classifier layer. In the offline training, the spatial and channel attention mechanism module is introduced to recalibrate the original features, and the space and channel weights are obtained respectively. The key features are selected by normalizing the weights to the corresponding original features. In online tracking, the network parameters of the full connection layer and classifier layer are trained, and the boundary box regression is used. Secondly, samples are collected according to the set threshold, and the negative sample with the highest classifier score is selected for each iteration to fine tune the network layer parameters. The experimental results on OTB2015 dataset show that compared with other mainstream tracking algorithms, the proposed method achieves better tracking accuracy. The overlap success rate and error success rate are 67.6% and 91.2% respectively.
Object tracking is one of the important research fields in computer vision. However, most tracking algorithm can not effectively learn the features suitable for tracking scene, which limits the performance improvement of tracking algorithm. To overcome this problem, this paper proposes a target tracking algorithm based on CNN Spatial and Channel Attention Mechanisms (CNNSCAM). The method consists of an off-line training apparent model and an adaptive updating classifier layer. In the offline training, the spatial and channel attention mechanism module is introduced to recalibrate the original features, and the space and channel weights are obtained respectively. The key features are selected by normalizing the weights to the corresponding original features. In online tracking, the network parameters of the full connection layer and classifier layer are trained, and the boundary box regression is used. Secondly, samples are collected according to the set threshold, and the negative sample with the highest classifier score is selected for each iteration to fine tune the network layer parameters. The experimental results on OTB2015 dataset show that compared with other mainstream tracking algorithms, the proposed method achieves better tracking accuracy. The overlap success rate and error success rate are 67.6% and 91.2% respectively.
2021, 43(9): 2577-2585.
doi: 10.11999/JEIT200362
Abstract:
Considering that previous tiny face hallucination methods either produced visually less pleasant faces or required architecturally more complex networks, this paper advocates a new deep model for tiny face hallucination by borrowing the idea of Relativistic Generative Adversarial Network (tfh-RGAN). Specifically, a hallucination generator and a relativistic discriminator are jointly learned in an alternately iterative training fashion by minimizing the combined pixel loss and relativistic generative adversarial loss. As for the generator, it is mainly structured as concatenation of a few basic modules followed by three 2×up-sampling layers, and each basic module is formulated by coupling the residual blocks, dense blocks, and depthwise separable convolution operators. As such, the generator can be made lightweight while with a considerable depth so as to achieve high quality face hallucination. As for the discriminator, it makes use of VGG128 while removing all its batch normalization layers and embedding a fully connected layer additionally so as to fulfill the capacity limit of relativistic adversarial learning. Experimental results reveal that, the proposed method, though simpler in the network architecture without a need of explicitly imposing any face structural prior, is able to produce better hallucination faces with higher definition and stronger reality. In terms of the quantitative assessment, the peak signal-to-noise ratio of the proposed method can be improved up to 0.25～1.51 dB compared against several previous approaches.
Considering that previous tiny face hallucination methods either produced visually less pleasant faces or required architecturally more complex networks, this paper advocates a new deep model for tiny face hallucination by borrowing the idea of Relativistic Generative Adversarial Network (tfh-RGAN). Specifically, a hallucination generator and a relativistic discriminator are jointly learned in an alternately iterative training fashion by minimizing the combined pixel loss and relativistic generative adversarial loss. As for the generator, it is mainly structured as concatenation of a few basic modules followed by three 2×up-sampling layers, and each basic module is formulated by coupling the residual blocks, dense blocks, and depthwise separable convolution operators. As such, the generator can be made lightweight while with a considerable depth so as to achieve high quality face hallucination. As for the discriminator, it makes use of VGG128 while removing all its batch normalization layers and embedding a fully connected layer additionally so as to fulfill the capacity limit of relativistic adversarial learning. Experimental results reveal that, the proposed method, though simpler in the network architecture without a need of explicitly imposing any face structural prior, is able to produce better hallucination faces with higher definition and stronger reality. In terms of the quantitative assessment, the peak signal-to-noise ratio of the proposed method can be improved up to 0.25～1.51 dB compared against several previous approaches.
2021, 43(9): 2586-2593.
doi: 10.11999/JEIT200884
Abstract:
The Kernel Extreme Learning Machine (KELM) with convex optimization form has higher classification accuracy, but it takes longer time to train kelm with iterative method than solving linear equation method of traditional kelm. To solve this problem, an Alternating Direction Multiplier Method(ADMM) of Binary Splitting (BSADMM-KELM) is proposed to improve the training speed of convex optimization kernel extreme learning machine. Firstly, the process of finding the optimal solution of the kernel extreme learning machine is split into two intermediate operators by introducing a binary splitting operator, and then the optimal solution of the original problem is obtained through the iterative calculation of the intermediate operators. On 22 UCI datasets, the training time of the proposed algorithm is 29 times faster than that of the effective set method and 4 times faster than that of the interior point method. The classification accuracy of the proposed algorithm is also better than that of the traditional kernel extreme learning machine. On large-scale datasets, the training time of the proposed algorithm is better than that of the traditional kernel extreme learning machine.
The Kernel Extreme Learning Machine (KELM) with convex optimization form has higher classification accuracy, but it takes longer time to train kelm with iterative method than solving linear equation method of traditional kelm. To solve this problem, an Alternating Direction Multiplier Method(ADMM) of Binary Splitting (BSADMM-KELM) is proposed to improve the training speed of convex optimization kernel extreme learning machine. Firstly, the process of finding the optimal solution of the kernel extreme learning machine is split into two intermediate operators by introducing a binary splitting operator, and then the optimal solution of the original problem is obtained through the iterative calculation of the intermediate operators. On 22 UCI datasets, the training time of the proposed algorithm is 29 times faster than that of the effective set method and 4 times faster than that of the interior point method. The classification accuracy of the proposed algorithm is also better than that of the traditional kernel extreme learning machine. On large-scale datasets, the training time of the proposed algorithm is better than that of the traditional kernel extreme learning machine.
2021, 43(9): 2594-2600.
doi: 10.11999/JEIT200882
Abstract:
For the parameters identification of Wiener nonlinear time-varying systems, iterative learning algorithms based on repeated axes are proposed to estimate the time-varying or even abrupt parameters. At first, the output nonlinear part of the Wiener system undertaken is tackled based on polynomial expansion, and then the regression model is constructed, the unknown parameters and intermediate variables are replaced by their estimates. Both iterative learning gradient and iterative learning least square algorithms are used to conduct the identification of the time-varying systems. Compared with the recursive algorithm with forgetting factor and iterative learning gradient algorithm, the simulation results demonstrate that the iterative learning least squares algorithm can perform high identification accuracy and efficiency, being of fast convergence speed and less resultant system output error, which verifies the effectiveness of the proposed algorithm.
For the parameters identification of Wiener nonlinear time-varying systems, iterative learning algorithms based on repeated axes are proposed to estimate the time-varying or even abrupt parameters. At first, the output nonlinear part of the Wiener system undertaken is tackled based on polynomial expansion, and then the regression model is constructed, the unknown parameters and intermediate variables are replaced by their estimates. Both iterative learning gradient and iterative learning least square algorithms are used to conduct the identification of the time-varying systems. Compared with the recursive algorithm with forgetting factor and iterative learning gradient algorithm, the simulation results demonstrate that the iterative learning least squares algorithm can perform high identification accuracy and efficiency, being of fast convergence speed and less resultant system output error, which verifies the effectiveness of the proposed algorithm.
2021, 43(9): 2601-2607.
doi: 10.11999/JEIT200725
Abstract:
In order to solve the problem that it is difficult to extract the Unmanned Aerial Vehicle (UAV) positioning signal from the environment with severe multipath interference in the passive positioning of the UAV, a UAV positioning signal separation based on Support Vector Machines (SVM) algorithm is proposed. During the training of the SVM model, the information entropy is obtained by calculating the Euclidean distance between the adjacent data sets of the UAV, and the model data is provided for the SVM to map the high-dimensional space. On this basis, the soft boundary of the threshold of the mapping function is added to make the model have the ability to adjust parameters adaptively to adapt to the data difference caused by the flexible movement of the UAV. Finally, an observer operating characteristic curve is constructed to obtain the result of UAV positioning signal separation. The simulation results show that the proposed algorithm can effectively separate the UAV positioning signal and noise.
In order to solve the problem that it is difficult to extract the Unmanned Aerial Vehicle (UAV) positioning signal from the environment with severe multipath interference in the passive positioning of the UAV, a UAV positioning signal separation based on Support Vector Machines (SVM) algorithm is proposed. During the training of the SVM model, the information entropy is obtained by calculating the Euclidean distance between the adjacent data sets of the UAV, and the model data is provided for the SVM to map the high-dimensional space. On this basis, the soft boundary of the threshold of the mapping function is added to make the model have the ability to adjust parameters adaptively to adapt to the data difference caused by the flexible movement of the UAV. Finally, an observer operating characteristic curve is constructed to obtain the result of UAV positioning signal separation. The simulation results show that the proposed algorithm can effectively separate the UAV positioning signal and noise.
2021, 43(9): 2608-2616.
doi: 10.11999/JEIT200583
Abstract:
In the large-scale and heterogeneous network environment, the industrial Internet service has the characteristics of small scale and certainty, so it is difficult to match the resources of the heterogeneous bearer network with the scheduling of the service and the orchestration of the function chain. Based on this, a multi-node joint scheduling model based on Non-workconserving is proposed. First, the full-path time coordination algorithm is adopted to extend the function chain from the spatial dimension topological orchestration to the space-time dimension; Then, aiming at the problem of synchronous scheduling in network nodes, a flow scheduling algorithm based on urgency is proposed to smooth delay jittering, furthermore, time-triggered scheduling is extended to large-scale, heterogeneous and non-synchronous bearer networks. A virtual arrival queue scheduling algorithm is proposed, which uses service synchronization mechanism instead of time synchronization to ensure the reachability requirements of service determination. Simulation results show that the algorithm proposed in this paper can improve the accessibility of the service and ensure service can reach in time, on time or cooperatively.
In the large-scale and heterogeneous network environment, the industrial Internet service has the characteristics of small scale and certainty, so it is difficult to match the resources of the heterogeneous bearer network with the scheduling of the service and the orchestration of the function chain. Based on this, a multi-node joint scheduling model based on Non-workconserving is proposed. First, the full-path time coordination algorithm is adopted to extend the function chain from the spatial dimension topological orchestration to the space-time dimension; Then, aiming at the problem of synchronous scheduling in network nodes, a flow scheduling algorithm based on urgency is proposed to smooth delay jittering, furthermore, time-triggered scheduling is extended to large-scale, heterogeneous and non-synchronous bearer networks. A virtual arrival queue scheduling algorithm is proposed, which uses service synchronization mechanism instead of time synchronization to ensure the reachability requirements of service determination. Simulation results show that the algorithm proposed in this paper can improve the accessibility of the service and ensure service can reach in time, on time or cooperatively.
2021, 43(9): 2617-2623.
doi: 10.11999/JEIT200231
Abstract:
In Software Defined Networks (SDN), in order to meet various network performance optimization goals, the control plane needs to frequently update the data plane. However, due to the asynchronous nature of the data plane, unreasonable updates will severely degrade network performance. To address this issue, a Fast and Consistent Flow Update (FCFU) strategy is proposed, which weakens the original strong dependency through flow segmentations and enables parallel updates. By analyzing the dependency relationship between sub-flow segments and multiple resources, the update schedule with less numbers of rounds is obtained. Finally, consistent flow update is achieved based on the delay queue. Experimental results show that, compared with the existing flow update algorithms, this strategy can shorten the total completion time of flow update by 20.6%, while ensuring that no congestion and packet reordering occur during the update period.
In Software Defined Networks (SDN), in order to meet various network performance optimization goals, the control plane needs to frequently update the data plane. However, due to the asynchronous nature of the data plane, unreasonable updates will severely degrade network performance. To address this issue, a Fast and Consistent Flow Update (FCFU) strategy is proposed, which weakens the original strong dependency through flow segmentations and enables parallel updates. By analyzing the dependency relationship between sub-flow segments and multiple resources, the update schedule with less numbers of rounds is obtained. Finally, consistent flow update is achieved based on the delay queue. Experimental results show that, compared with the existing flow update algorithms, this strategy can shorten the total completion time of flow update by 20.6%, while ensuring that no congestion and packet reordering occur during the update period.
2021, 43(9): 2624-2631.
doi: 10.11999/JEIT200198
Abstract:
Compared with the Wi-Fi Received Signal Strength (RSS) commonly-used for the indoor localization, the Channel State Information (CSI) can be used for the precise ranging to achieve the high Wi-Fi indoor localization accuracy since it includes the fine-grained physical-layer information such as the amplitude and phase of each subcarrier during the signal transmission. Due to the lack of theoretical analysis of the localization error bound in existing CSI ranging-based localization methods, it is difficult to compare the ideal performance of different localization methods. Therefore, a CSI ranging-based Wi-Fi indoor localization error bound analysis method is proposed, which is based on the indoor signal propagation model to derive out the CSI ranging-based localization error bound with the clock asynchronous effect by considering the relationship between the localization accuracy and the path loss, shadow fading, and multipath effect. Besides, through the experimental comparison, this paper analyzes the difference between the actual localization error and the derived localization error bound, as well as discusses the impact of different experimental parameters on the localization performance.
Compared with the Wi-Fi Received Signal Strength (RSS) commonly-used for the indoor localization, the Channel State Information (CSI) can be used for the precise ranging to achieve the high Wi-Fi indoor localization accuracy since it includes the fine-grained physical-layer information such as the amplitude and phase of each subcarrier during the signal transmission. Due to the lack of theoretical analysis of the localization error bound in existing CSI ranging-based localization methods, it is difficult to compare the ideal performance of different localization methods. Therefore, a CSI ranging-based Wi-Fi indoor localization error bound analysis method is proposed, which is based on the indoor signal propagation model to derive out the CSI ranging-based localization error bound with the clock asynchronous effect by considering the relationship between the localization accuracy and the path loss, shadow fading, and multipath effect. Besides, through the experimental comparison, this paper analyzes the difference between the actual localization error and the derived localization error bound, as well as discusses the impact of different experimental parameters on the localization performance.
2021, 43(9): 2632-2639.
doi: 10.11999/JEIT200443
Abstract:
In view of the hotspot scenarios in B5G/6G, in order to meet the needs of its ultra-large network capacity, a multi-Unmanned Aerial Vehicle (UAV) assisted millimeter wave heterogeneous network model is constructed. In this network model, the distribution of Ground Base Station (G-BS) is modeled as Poisson point process, the distribution of UAV is modeled as Poisson cluster process, and the projection of the UAV on ground and the Ground User Equipment (GUE) are distributed around the G-BS. For the sake of exploring the contribution of inter-cluster association and the impact of inter-cluster interference, the 2-tier network model is extended to 4-tier network model composed of inter-cluster and intra-cluster Base Station (BS). And the 4-tier association scheme in which the GUE is associated with intra-cluster BS and inter-cluster BS at the same time is proposed. Initially, the path loss of each tier’s association distance is analyzed through the propagation model. Furthermore, using stochastic geometry method, combined with the interference of GUE in the downlink, the Signal-to-Interference plus Noise Ratio (SINR) coverage probability expression of GUE is derived. Finally, the simulation results show that the height of UAV and the average number of cluster members have non-monotonic effect on SINR coverage probability. When UAV height is low, compared with the 2-tier association scheme that GUE is only associated with the intra-cluster BS, the 4-tier association scheme proposed in this paper can improve the SINR coverage probability significantly.
In view of the hotspot scenarios in B5G/6G, in order to meet the needs of its ultra-large network capacity, a multi-Unmanned Aerial Vehicle (UAV) assisted millimeter wave heterogeneous network model is constructed. In this network model, the distribution of Ground Base Station (G-BS) is modeled as Poisson point process, the distribution of UAV is modeled as Poisson cluster process, and the projection of the UAV on ground and the Ground User Equipment (GUE) are distributed around the G-BS. For the sake of exploring the contribution of inter-cluster association and the impact of inter-cluster interference, the 2-tier network model is extended to 4-tier network model composed of inter-cluster and intra-cluster Base Station (BS). And the 4-tier association scheme in which the GUE is associated with intra-cluster BS and inter-cluster BS at the same time is proposed. Initially, the path loss of each tier’s association distance is analyzed through the propagation model. Furthermore, using stochastic geometry method, combined with the interference of GUE in the downlink, the Signal-to-Interference plus Noise Ratio (SINR) coverage probability expression of GUE is derived. Finally, the simulation results show that the height of UAV and the average number of cluster members have non-monotonic effect on SINR coverage probability. When UAV height is low, compared with the 2-tier association scheme that GUE is only associated with the intra-cluster BS, the 4-tier association scheme proposed in this paper can improve the SINR coverage probability significantly.
2021, 43(9): 2640-2646.
doi: 10.11999/JEIT200248
Abstract:
Multi-Mode Orthogonal Frequency Division Multiplexing system with Index Modulation(MM-OFDM-IM) uses different constellation sets on the basis of orthogonal frequency division Multiplexing with index modulation to index modulate all subcarriers in the system, which can effectively improve the subcarrier utilization and spectrum efficiency. However, the utilization of all subcarriers will affect the anti-interference ability of the system sub-carriers, which will cause the system’s bit error rate performance to decrease. To solve this problem, a Permutation-Mode Orthogonal Frequency Division Multiplexing with Index Modulation (PM-OFDM-IM) is proposed, which is based on the MM-OFDM-IM. This system re-introduces the silent subcarrier, which can ensure the system’s higher spectral efficiency and improve the system’s bit error rate performance. Further, a classification mapping mode is proposed based on amplitude phase shift keying, i.e. a Permutation Constellation set Classification mode arranged by Radius (PCC-R). This mode can combine system information well. Finally, simulation results verify that the system can better balance the spectral efficiency and the bit error rate performance of the system, and the proposed classification mapping scheme can achieve better system performance.
Multi-Mode Orthogonal Frequency Division Multiplexing system with Index Modulation(MM-OFDM-IM) uses different constellation sets on the basis of orthogonal frequency division Multiplexing with index modulation to index modulate all subcarriers in the system, which can effectively improve the subcarrier utilization and spectrum efficiency. However, the utilization of all subcarriers will affect the anti-interference ability of the system sub-carriers, which will cause the system’s bit error rate performance to decrease. To solve this problem, a Permutation-Mode Orthogonal Frequency Division Multiplexing with Index Modulation (PM-OFDM-IM) is proposed, which is based on the MM-OFDM-IM. This system re-introduces the silent subcarrier, which can ensure the system’s higher spectral efficiency and improve the system’s bit error rate performance. Further, a classification mapping mode is proposed based on amplitude phase shift keying, i.e. a Permutation Constellation set Classification mode arranged by Radius (PCC-R). This mode can combine system information well. Finally, simulation results verify that the system can better balance the spectral efficiency and the bit error rate performance of the system, and the proposed classification mapping scheme can achieve better system performance.
2021, 43(9): 2647-2653.
doi: 10.11999/JEIT200441
Abstract:
Hybrid precoding is essential to improve the performance of multi-user millimeter-wave massive Multiple Input Multiple Output (MIMO) systems, but currently hybrid precoding based on fully connected structures and sub-connected structures have the problems of high energy consumption and severe performance loss, respectively. This paper comprehensively considers the spectral efficiency and energy efficiency of the system, proposes hybrid dynamic connection structure, and designs hybrid precoding algorithm under the structure. In this algorithm, the analog domain precoding of the hybrid dynamic connection structure is designed by maximizing the increment of Signal-to-Interference-plus-Noise Ratio (SINR), and then the digital domain precoding is designed by Block Diagonalization (BD) to suppress multi-user multi-stream interference through an equivalent channel. Simulation experiments show that the spectral efficiency of the proposed hybrid dynamic connection structure is between the spectral efficiency of the fully connected structure and the spectral efficiency of the hybrid fixed connection structure, but the highest energy efficiency is obtained.
Hybrid precoding is essential to improve the performance of multi-user millimeter-wave massive Multiple Input Multiple Output (MIMO) systems, but currently hybrid precoding based on fully connected structures and sub-connected structures have the problems of high energy consumption and severe performance loss, respectively. This paper comprehensively considers the spectral efficiency and energy efficiency of the system, proposes hybrid dynamic connection structure, and designs hybrid precoding algorithm under the structure. In this algorithm, the analog domain precoding of the hybrid dynamic connection structure is designed by maximizing the increment of Signal-to-Interference-plus-Noise Ratio (SINR), and then the digital domain precoding is designed by Block Diagonalization (BD) to suppress multi-user multi-stream interference through an equivalent channel. Simulation experiments show that the spectral efficiency of the proposed hybrid dynamic connection structure is between the spectral efficiency of the fully connected structure and the spectral efficiency of the hybrid fixed connection structure, but the highest energy efficiency is obtained.
2021, 43(9): 2654-2662.
doi: 10.11999/JEIT200256
Abstract:
To optimize strategy of resource allocation and task offloading decision on D2D-assisted cloud-fog architecture, a joint resource allocation and offloading decision algorithm based on a multi-agent architecture deep reinforcement learning method is proposed. Firstly, considering incentive constraints, energy constraints, and network resource constraints, the algorithm jointly optimizes wireless resource allocation, computing resource allocation, and offloading decisions. Further, the algorithm establishes a stochastic optimization model that maximizes the total user Quality of Experience (QoE) of the system, and transfers it into an MDP problem. Secondly, the algorithm factorizes the original MDP problem and models a Markov game. Then, a centralized training and distributed execution mechanism based on the Actor-Critic (AC) algorithm is proposed. In the centralized training process, multi-agents obtains the global information through cooperation to optimize the resource allocation and task offloading decision strategies. After the training process, each agent performs independently resource allocation and task offloading based on the current system state and strategy. Finally, the simulation results demonstrate that the algorithm can effectively improve user QoE, and reduce delay and energy consumption.
To optimize strategy of resource allocation and task offloading decision on D2D-assisted cloud-fog architecture, a joint resource allocation and offloading decision algorithm based on a multi-agent architecture deep reinforcement learning method is proposed. Firstly, considering incentive constraints, energy constraints, and network resource constraints, the algorithm jointly optimizes wireless resource allocation, computing resource allocation, and offloading decisions. Further, the algorithm establishes a stochastic optimization model that maximizes the total user Quality of Experience (QoE) of the system, and transfers it into an MDP problem. Secondly, the algorithm factorizes the original MDP problem and models a Markov game. Then, a centralized training and distributed execution mechanism based on the Actor-Critic (AC) algorithm is proposed. In the centralized training process, multi-agents obtains the global information through cooperation to optimize the resource allocation and task offloading decision strategies. After the training process, each agent performs independently resource allocation and task offloading based on the current system state and strategy. Finally, the simulation results demonstrate that the algorithm can effectively improve user QoE, and reduce delay and energy consumption.
2021, 43(9): 2663-2671.
doi: 10.11999/JEIT200473
Abstract:
In view of the widely exist co-channel interference in D2D communication system, a novel method to quantify the co-channel interference based on NonDate Aided Error Vector Magnitude (NDA-EVM) is proposed. NDA-EVM is considered as a new metric to evaluate the change of the channels. The NDA-EVM of M-QAM modulated signal is analytically derived. Moreover, the model of NDA-EVM under co-channel interference is established. Specifically, the upper bound of NDA-EVM is calculated when co-channel interferences exist, so co-channel interferences can be quantified. Theoretical analysis and simulation experiments indicate that, when compared with traditional NDA-EVM algorithm, the proposed upper bound of NDA-EVM reduces the time complexity of algorithm from O(M2) to
In view of the widely exist co-channel interference in D2D communication system, a novel method to quantify the co-channel interference based on NonDate Aided Error Vector Magnitude (NDA-EVM) is proposed. NDA-EVM is considered as a new metric to evaluate the change of the channels. The NDA-EVM of M-QAM modulated signal is analytically derived. Moreover, the model of NDA-EVM under co-channel interference is established. Specifically, the upper bound of NDA-EVM is calculated when co-channel interferences exist, so co-channel interferences can be quantified. Theoretical analysis and simulation experiments indicate that, when compared with traditional NDA-EVM algorithm, the proposed upper bound of NDA-EVM reduces the time complexity of algorithm from O(M2) to
2021, 43(9): 2672-2679.
doi: 10.11999/JEIT200849
Abstract:
The load imbalance in the wireless domain limits the overall transmission efficiency of the network in the joint architecture of Time and Wavelength Division Multiplexed Passive Optical Network (TWDM-PON) and Cloud Radio Access Network (C-RAN). A Load Balancing User Association and Resource Allocation (LBUARA) algorithm is proposed to ensure the Quality of Service(QoS) of users, and make full use of network resources TWDM-PON jointly with C-RAN architecture. Firstly, the user revenue function is constructed according to the service quality requirements of different users and the impact of Remote Radio Head (RRH) load on users. Furthermore, a random game model is established according to the network state, under the premise of ensuring the quality of user service. A user association and resource allocation algorithm based on multi-agent Q-learning load balancing is proposed to obtain the optimal user association and resource allocation plan. The simulation results show that users association and resource allocation strategies mentioned can achieve load balancing network to ensure quality of service users, and improve network throughput.
The load imbalance in the wireless domain limits the overall transmission efficiency of the network in the joint architecture of Time and Wavelength Division Multiplexed Passive Optical Network (TWDM-PON) and Cloud Radio Access Network (C-RAN). A Load Balancing User Association and Resource Allocation (LBUARA) algorithm is proposed to ensure the Quality of Service(QoS) of users, and make full use of network resources TWDM-PON jointly with C-RAN architecture. Firstly, the user revenue function is constructed according to the service quality requirements of different users and the impact of Remote Radio Head (RRH) load on users. Furthermore, a random game model is established according to the network state, under the premise of ensuring the quality of user service. A user association and resource allocation algorithm based on multi-agent Q-learning load balancing is proposed to obtain the optimal user association and resource allocation plan. The simulation results show that users association and resource allocation strategies mentioned can achieve load balancing network to ensure quality of service users, and improve network throughput.
2021, 43(9): 2680-2687.
doi: 10.11999/JEIT200411
Abstract:
During the marine moving target detection for airborne early warning radar, the high-speed movement of the radar platform causes the serious broadening of the sea clutter Doppler spectrum, which affects the target detection performance. To solve this problem, a clutter suppression method called Space-Time Adaptive Processing (STAP) is effective, which exploits the space-time coupling characteristics of clutter. However, compared with the land clutter, the motion characteristics of sea clutter lead to the broadening of the clutter space-time spectrum, resulting in the clutter Doppler frequency and the spatial cone angle no longer maintaining a one-to-one correspondence; thus the clutter suppression performance significantly degrades. According to the motion characteristics of sea clutter, a robust subspace projection method is proposed in this paper. This method improves the robustness of clutter suppression by using the adaptive notch broadening technique and the filter then adapt technique. Finally, the effectiveness of this method is verified through the simulation results and the real-measured sea clutter data.
During the marine moving target detection for airborne early warning radar, the high-speed movement of the radar platform causes the serious broadening of the sea clutter Doppler spectrum, which affects the target detection performance. To solve this problem, a clutter suppression method called Space-Time Adaptive Processing (STAP) is effective, which exploits the space-time coupling characteristics of clutter. However, compared with the land clutter, the motion characteristics of sea clutter lead to the broadening of the clutter space-time spectrum, resulting in the clutter Doppler frequency and the spatial cone angle no longer maintaining a one-to-one correspondence; thus the clutter suppression performance significantly degrades. According to the motion characteristics of sea clutter, a robust subspace projection method is proposed in this paper. This method improves the robustness of clutter suppression by using the adaptive notch broadening technique and the filter then adapt technique. Finally, the effectiveness of this method is verified through the simulation results and the real-measured sea clutter data.
2021, 43(9): 2688-2694.
doi: 10.11999/JEIT200873
Abstract:
In view of the fact that low power resource utilization rate exists in radar network, a Target Capacity based Power Allocation (TC-PA) scheme is proposed to increase the number of the targets that satisfy tracking accuracy requirements. Firstly, this scheme formulates the power allocation model of radar network as a non-smooth and non-convex optimization problem. Then the original problem is relaxed into a smooth and non-convex problem through introducing Sigmoid function. Finally, the relaxed non-convex problem is solved by utilizing the Proximal Inexact Augmented Lagrangian Multiplier Method (PI-ALMM). Simulation results show that the PI-ALMM can quickly converge to a stationary point for solving the non-convex optimization problem with linear constraints. Moreover, the proposed TC-PA scheme outperforms the traditional uniform power allocation method and genetic algorithm, in terms of target capacity.
In view of the fact that low power resource utilization rate exists in radar network, a Target Capacity based Power Allocation (TC-PA) scheme is proposed to increase the number of the targets that satisfy tracking accuracy requirements. Firstly, this scheme formulates the power allocation model of radar network as a non-smooth and non-convex optimization problem. Then the original problem is relaxed into a smooth and non-convex problem through introducing Sigmoid function. Finally, the relaxed non-convex problem is solved by utilizing the Proximal Inexact Augmented Lagrangian Multiplier Method (PI-ALMM). Simulation results show that the PI-ALMM can quickly converge to a stationary point for solving the non-convex optimization problem with linear constraints. Moreover, the proposed TC-PA scheme outperforms the traditional uniform power allocation method and genetic algorithm, in terms of target capacity.
2021, 43(9): 2695-2702.
doi: 10.11999/JEIT200515
Abstract:
The radiation efficiency of large-sized ElectroMagnetic Vector Sensor (EMVS) composed of long electric-dipoles or large magnetic-loops is higher than that of small-sized EMVS. The study of its parameter estimation algorithm is helpful to promote the practical application of EMVS. To solve the problem of parameter estimation of coherent targets with sparse array of spatially spread long electric-dipoles, a high accuracy and unambiguous closed multi-dimensional parameter solution algorithm is proposed. First, the high accuracy and periodically ambiguous direction-cosine estimations are obtained by using the spatial rotation invariance and the internal attributes of a single vector sensor. Then, the two-dimensional direction of arrival coarse estimations are derived based on the steering vector of a single vector sensor. Finally, the high accuracy and unambiguous multi-dimensional parameter estimations are obtained with disambiguation method. This proposed algorithm avoids the loss of polarization information and iterative search process of the traditional polarization smoothing algorithm, and can realize automatic matching of parameters. The computer simulation results show the effectiveness of the proposed algorithm in decoherence of the separated long electric dipole array.
The radiation efficiency of large-sized ElectroMagnetic Vector Sensor (EMVS) composed of long electric-dipoles or large magnetic-loops is higher than that of small-sized EMVS. The study of its parameter estimation algorithm is helpful to promote the practical application of EMVS. To solve the problem of parameter estimation of coherent targets with sparse array of spatially spread long electric-dipoles, a high accuracy and unambiguous closed multi-dimensional parameter solution algorithm is proposed. First, the high accuracy and periodically ambiguous direction-cosine estimations are obtained by using the spatial rotation invariance and the internal attributes of a single vector sensor. Then, the two-dimensional direction of arrival coarse estimations are derived based on the steering vector of a single vector sensor. Finally, the high accuracy and unambiguous multi-dimensional parameter estimations are obtained with disambiguation method. This proposed algorithm avoids the loss of polarization information and iterative search process of the traditional polarization smoothing algorithm, and can realize automatic matching of parameters. The computer simulation results show the effectiveness of the proposed algorithm in decoherence of the separated long electric dipole array.
2021, 43(9): 2703-2710.
doi: 10.11999/JEIT200402
Abstract:
The matrix Constant False Alarm Rate (CFAR) detector based on the information geometry theory is an effective method for target detection in the K-distributed sea clutter environment. However, the general matrix CFAR method has a high computational complexity and its detection performance is not as good as Adaptive Normalized Matched Filter (ANMF) when the target Doppler frequency deviates from the clutter spectrum center seriously, which affects its practical application. For this reason, considered the filtered received signal by the filter bank, a Matrix CFAR Detection method based on the Filter bank subband Decomposition of Maximum Eigenvalue (FD-MEMD) is proposed. The double clutter suppression helps to solve the problem that Matrix CFAR is invalid when the target Doppler frequency is far away the central of the clutter spectrum. Finally, the simulation results show that the improved FD-MEMD has a good detection performance.
The matrix Constant False Alarm Rate (CFAR) detector based on the information geometry theory is an effective method for target detection in the K-distributed sea clutter environment. However, the general matrix CFAR method has a high computational complexity and its detection performance is not as good as Adaptive Normalized Matched Filter (ANMF) when the target Doppler frequency deviates from the clutter spectrum center seriously, which affects its practical application. For this reason, considered the filtered received signal by the filter bank, a Matrix CFAR Detection method based on the Filter bank subband Decomposition of Maximum Eigenvalue (FD-MEMD) is proposed. The double clutter suppression helps to solve the problem that Matrix CFAR is invalid when the target Doppler frequency is far away the central of the clutter spectrum. Finally, the simulation results show that the improved FD-MEMD has a good detection performance.
2021, 43(9): 2711-2719.
doi: 10.11999/JEIT200300
Abstract:
As it is difficult to balance the sparse and focusing features for conventional sparse autofocusing algorithm of Synthetic Aperture Radar (SAR), a Multi-task Learning Sparse Autofocusing (MtL-SA) algorithm is proposed under a novel Alternating Direction Method of Multipliers (ADMM) in this paper. The image entropy norm is introduced to model the focusing feature of the SAR imagery, and it is minimized in a regularized manner using the proximal algorithm. To overcome the non-convexity of the original objective function, a surrogate function under the ADMM framework is designed and optimized accordingly. This ensures closed-form solution of the errors and the focusing feature. Besides, the\begin{document}$ \ell {_1}$\end{document} ![]()
![]()
As it is difficult to balance the sparse and focusing features for conventional sparse autofocusing algorithm of Synthetic Aperture Radar (SAR), a Multi-task Learning Sparse Autofocusing (MtL-SA) algorithm is proposed under a novel Alternating Direction Method of Multipliers (ADMM) in this paper. The image entropy norm is introduced to model the focusing feature of the SAR imagery, and it is minimized in a regularized manner using the proximal algorithm. To overcome the non-convexity of the original objective function, a surrogate function under the ADMM framework is designed and optimized accordingly. This ensures closed-form solution of the errors and the focusing feature. Besides, the
2021, 43(9): 2720-2727.
doi: 10.11999/JEIT200692
Abstract:
The mutual coupling effects of the transmitter and receiver are known to degrade the performance of direction finding for a bistatic Multiple Input Multiple Output (MIMO) radar system. A four cumulant-combinatorial matrix-based algorithm is proposed to estimate jointly the Direction Of Departure (DOD)and Direction Of Arrival (DOA) of targets under the coexistences of unknown mutual coupling and Gaussian colored noise. Firstly, the multiple groups of four cumulant matrices both on transmitter and receiver are constructed by using the Kronecker product and the banded symmetric Toeplitz characteristics of the mutual coupling matrices. The block four-cumulant matrix is further constructed by combining the transmitter and receiver four cunmulant matrices. Then the new matrix is combined to extract the transmit and receive shift invariance matrices by using the transmitter and receiver four cumulant matrices. The results illustrate that: The proposed method can estimate the DOD and DOA of the targets efficiently in the presence of the strong mutual coupling effect, and parameters are paired automatically without extra pairing operation. The parameter estimation performance of the proposed method is better than those of the existing methods under the strong mutual coupling effect conditions.
The mutual coupling effects of the transmitter and receiver are known to degrade the performance of direction finding for a bistatic Multiple Input Multiple Output (MIMO) radar system. A four cumulant-combinatorial matrix-based algorithm is proposed to estimate jointly the Direction Of Departure (DOD)and Direction Of Arrival (DOA) of targets under the coexistences of unknown mutual coupling and Gaussian colored noise. Firstly, the multiple groups of four cumulant matrices both on transmitter and receiver are constructed by using the Kronecker product and the banded symmetric Toeplitz characteristics of the mutual coupling matrices. The block four-cumulant matrix is further constructed by combining the transmitter and receiver four cunmulant matrices. Then the new matrix is combined to extract the transmit and receive shift invariance matrices by using the transmitter and receiver four cumulant matrices. The results illustrate that: The proposed method can estimate the DOD and DOA of the targets efficiently in the presence of the strong mutual coupling effect, and parameters are paired automatically without extra pairing operation. The parameter estimation performance of the proposed method is better than those of the existing methods under the strong mutual coupling effect conditions.
2021, 43(9): 2728-2734.
doi: 10.11999/JEIT190745
Abstract:
ISAR resolution evaluation is an important part of accuracy evaluation of space target imaging radar. Two difficult problems, i.e. the selection of evaluation criteria and the design of evaluation methods are discussed in detail. Then an ISAR cross-range scaling algorithm based on high precision orbit and telemetry attitude of space target is proposed. A new ISAR resolution evaluation method is established based on the new scaling technique. The method is validated by using the actual imaging test data for different types of space targets.
ISAR resolution evaluation is an important part of accuracy evaluation of space target imaging radar. Two difficult problems, i.e. the selection of evaluation criteria and the design of evaluation methods are discussed in detail. Then an ISAR cross-range scaling algorithm based on high precision orbit and telemetry attitude of space target is proposed. A new ISAR resolution evaluation method is established based on the new scaling technique. The method is validated by using the actual imaging test data for different types of space targets.
2021, 43(9): 2735-2742.
doi: 10.11999/JEIT200888
Abstract:
Considering the problem of obtaining the reference signal for passive radar with Orthogonal Frequency Division Multiplexing (OFDM) waveform, the reconstruction method based on "demodulation-remodulation" employs the waveform advantage to obtain a purer reference signal. On this basis, a Deep Neural Network (DNN) reconstruction method that combines OFDM demodulation, channel estimation, channel equalization, and constellation point inverse mapping is proposed to establish a DNN-based reference signal reconstruction scheme. This method can be used to adaptively and deeply excavate the mapping relationship between time-domain received symbols and transmission symbols through network learning, and implicitly estimate the channel response, thereby improving demodulation accuracy and reconstruction performance. Firstly, the acquisition of simulation data sets, the construction and training of DNN are studied in this paper.Then, the comparison between the DNN method and the traditional method about reference signal reconstruction performance is analyzed under the condition that the number of pilots is reduced, the cyclic prefix is removed, the symbol timing offset exists, the carrier frequency offset exists, the time domain windowing filter is performed on the high peak-to-average power ratio signal, and all the above parameters are superimposed. Finally, simulation results show the effectiveness of this method.
Considering the problem of obtaining the reference signal for passive radar with Orthogonal Frequency Division Multiplexing (OFDM) waveform, the reconstruction method based on "demodulation-remodulation" employs the waveform advantage to obtain a purer reference signal. On this basis, a Deep Neural Network (DNN) reconstruction method that combines OFDM demodulation, channel estimation, channel equalization, and constellation point inverse mapping is proposed to establish a DNN-based reference signal reconstruction scheme. This method can be used to adaptively and deeply excavate the mapping relationship between time-domain received symbols and transmission symbols through network learning, and implicitly estimate the channel response, thereby improving demodulation accuracy and reconstruction performance. Firstly, the acquisition of simulation data sets, the construction and training of DNN are studied in this paper.Then, the comparison between the DNN method and the traditional method about reference signal reconstruction performance is analyzed under the condition that the number of pilots is reduced, the cyclic prefix is removed, the symbol timing offset exists, the carrier frequency offset exists, the time domain windowing filter is performed on the high peak-to-average power ratio signal, and all the above parameters are superimposed. Finally, simulation results show the effectiveness of this method.
2021, 43(9): 2743-2750.
doi: 10.11999/JEIT200894
Abstract:
For the most of the existing gesture recognition methods based on the radar sensor, the parameters such as the distance, Doppler, and angle are estimated using the radar echo at first. And then the obtained data spectra are inputted into the convolutional neural networks to classify the gestures. The implementation process is complicated. A dynamic gesture recognition method is proposed based on the millimeter-wave radar using the One-Dimensional Series connection Neural Networks (1D-ScNN) in this paper. Firstly, the original echo of dynamic gesture is obtained by the millimeter-wave radar. The gesture features are extracted by the one-dimensional convolution and pooling operations, and then are inputted into the one-dimensional inception v3 structure. In order to aggregate the one-dimensional features, the Long Short-Term Memory (LSTM) modular is connected to the end of the network. The inter-frame correlation of dynamic gestures echo is fully utilized to improve the recognition accuracy and the convergence speed of training. The experimental results show that the proposed method is simple in implementation and has a fast convergence speed. The classification accuracy can reach more than 96.0%, which is higher than the traditional gesture classification methods.
For the most of the existing gesture recognition methods based on the radar sensor, the parameters such as the distance, Doppler, and angle are estimated using the radar echo at first. And then the obtained data spectra are inputted into the convolutional neural networks to classify the gestures. The implementation process is complicated. A dynamic gesture recognition method is proposed based on the millimeter-wave radar using the One-Dimensional Series connection Neural Networks (1D-ScNN) in this paper. Firstly, the original echo of dynamic gesture is obtained by the millimeter-wave radar. The gesture features are extracted by the one-dimensional convolution and pooling operations, and then are inputted into the one-dimensional inception v3 structure. In order to aggregate the one-dimensional features, the Long Short-Term Memory (LSTM) modular is connected to the end of the network. The inter-frame correlation of dynamic gestures echo is fully utilized to improve the recognition accuracy and the convergence speed of training. The experimental results show that the proposed method is simple in implementation and has a fast convergence speed. The classification accuracy can reach more than 96.0%, which is higher than the traditional gesture classification methods.
2021, 43(9): 2751-2756.
doi: 10.11999/JEIT200846
Abstract:
As one of the commonly used materials for microwave vacuum electronic devices, the evaporation characteristics of oxygen free copper materials will affect the electrical properties of microwave vacuum electronic devices. In this paper, the effect of treatment process on the evaporation performance of oxygen free copper is studied by using ultra-high vacuum testing equipment. The thickness of evaporated copper film is measured by X-ray thickness gauge. And the surface morphology of oxygen free copper is observed by Scanning Electron Microscope (SEM). The results show that the macro-appearance surface roughness has little effect on the evaporation performance of oxygen free copper materials, but the treatment process has a great influence on the evaporation performance. The evaporation capacity of oxygen free copper will be increased after acid pickling and the evaporation capacity of oxygen free copper materials can be reduced by calcination in hydrogen. The evaporation capacity of oxygen free copper treated by deoiling cleaning and calcination in hydrogen is very low. The surface of oxygen free copper is analyzed. It is found that the vacuum evaporation performance of oxygen free copper material is related to the surface morphology of the material. The microscopic surface is smooth without oxidation and holes, and the vacuum evaporation of oxygen free copper material is less.
As one of the commonly used materials for microwave vacuum electronic devices, the evaporation characteristics of oxygen free copper materials will affect the electrical properties of microwave vacuum electronic devices. In this paper, the effect of treatment process on the evaporation performance of oxygen free copper is studied by using ultra-high vacuum testing equipment. The thickness of evaporated copper film is measured by X-ray thickness gauge. And the surface morphology of oxygen free copper is observed by Scanning Electron Microscope (SEM). The results show that the macro-appearance surface roughness has little effect on the evaporation performance of oxygen free copper materials, but the treatment process has a great influence on the evaporation performance. The evaporation capacity of oxygen free copper will be increased after acid pickling and the evaporation capacity of oxygen free copper materials can be reduced by calcination in hydrogen. The evaporation capacity of oxygen free copper treated by deoiling cleaning and calcination in hydrogen is very low. The surface of oxygen free copper is analyzed. It is found that the vacuum evaporation performance of oxygen free copper material is related to the surface morphology of the material. The microscopic surface is smooth without oxidation and holes, and the vacuum evaporation of oxygen free copper material is less.
