面向智能家居的WiFi多链路协同人体跟踪方法

潘厚丞; 蔡宇双; 姚俊梅; 张霆廷

doi:10.11999/JEIT260267

面向智能家居的WiFi多链路协同人体跟踪方法

doi: 10.11999/JEIT260267 cstr: 32379.14.JEIT260267

潘厚丞^{1, 2},
蔡宇双^{1, 2},
姚俊梅³,
张霆廷^{1, 2, ,}

1.
哈尔滨工业大学(深圳)信息科学与技术学院深圳 518055
2.
广东省空天网络与智能感知重点实验室深圳 518055
3.
深圳大学计算机与软件学院深圳 518060

基金项目: 国家自然科学基金面上项目(62171160, 62027802)，哈尔滨工业大学原创前沿探索基金(HIT.OCEF.2022055)，深圳市重大科技专项(KJZD20231023093055002, KJZD20230923114804009)，广东省重点实验室(2024) (2024KSYS023)

详细信息

作者简介:
潘厚丞：男，硕士生，研究方向为信号处理、通信感知一体化等

蔡宇双：男，博士生，研究方向为Wi-Fi感知、多传感器融合、多视角学习等

姚俊梅：女，副教授，研究方向为无线网络优化、通信感知一体化、移动计算等

张霆廷：男，教授，研究方向为多智能体协同理论与实现、无线定位理论和脉冲超宽带技术等

通讯作者:
张霆廷　zhangtt@hit.edu.cn

中图分类号: TN929.5
计量
- 文章访问数: 29
- HTML全文浏览量: 13
- PDF下载量: 1
- 被引次数: 0
出版历程
- 收稿日期: 2026-03-11
- 修回日期: 2026-07-01
- 录用日期: 2026-07-01
- 网络出版日期: 2026-07-12

A WiFi Multi-Link Collaborative Human Tracking Method for Smart Home

PAN Houcheng^{1, 2},
CAI Yushuang^{1, 2},
YAO Junmei³,
ZHANG Tingting^{1, 2
, ,}

1.
School of Information Science and Technology, Harbin Institute of Technology (Shenzhen), Shenzhen 518055, China
2.
Guangdong Provincial Key Laboratory of Space-Aerial Networking and Intelligent Sensing, Shenzhen 518055, China
3.
College of Computer Science and Software Engineering, Shenzhen University, Shenzhen 518060, China

Funds: The Natural Science Foundation of China (62171160, 62027802), The Fundamental Research Funds for the Central Universities (HIT.OCEF.2022055), The Shenzhen Science and Technology Program (KJZD20231023093055002, KJZD20230923114804009), Guangdong Provincial Key Laboratory (2024) (2024KSYS023)

摘要

摘要: 在智能家居场景中，WiFi网络将多个物联网(IoT)设备相连，为实现室内被动人体跟踪提供了新路径。然而，多链路跟踪需要考虑信号融合、设备自定位等问题。为此，该文提出了一种适用于商用WiFi的多链路协同跟踪方法。针对商用网卡的测量信号噪声，这种方法首先进行信号预处理，从高维的原始信道状态信息(CSI)数据中提取多普勒频移(DFS)。随后，对多链路DFS进行融合，以实现协同跟踪。该方法在没有WiFi设备位置先验知识的条件下也能实现设备自定位，为跟踪算法提供输入。最后，该文搭建了一个多发单收网络结构的原型验证系统，在未知设备位置的情况下实现了小于50 cm的人体跟踪精度，相比传统方法提升了约70%。在复杂场景的测试中表明，该系统能达到亚米级的跟踪精度，对设备位置误差具有鲁棒性，并显示出一定实时运行潜力。
- WiFi感知 /
- 多链路 /
- 被动人体跟踪 /
- 多普勒频移
Abstract: Objective WiFi-based indoor passive human tracking has become increasingly attractive for smart homes due to the widespread interconnection of Internet of Things (IoT) devices via existing WiFi infrastructures. Some approaches estimate Angle of Arrival (AoA) or Time of Flight (ToF) of the target reflection path to achieve tracking, yet which are fundamentally limited by the difficulty of distinguishing weak target reflection component from dense multipath due to small antenna array and narrow bandwidth under the existing mature protocols. Consequently, Doppler Frequency Shift (DFS)-based approaches utilizing multiple links have proven more practical. Dead Reckoning (DR) is commonly applied in such methods. However, it is challenging to dynamically select links for signal fusion. Furthermore, the performance of DR-based methods is significantly degraded when device locations are inaccurate. Additionally, Existing methods usually adopt a single-transmitter-multi-receiver structure, which complicates data aggregation and restricts multi-antenna utilization. To address the challenges of multi-link signal fusion and device-location uncertainty, Particle Filter (PF) is used in this paper. Furthermore, a multi-transmitter-single-receiver structure is adopted, in which multiple STAtions (STAs) transmit uplink packets to a single Access Point (AP). This structure simplifies data aggregation while exploiting the multi-antenna advantage at the AP. Methods In the IEEE 802.11 standard, the wireless channel is estimated at the receiver using pilots embedded in WiFi packets, and Channel State Information (CSI) is continuously reported. Because human motion perturbs multipath propagation and causes time-varying changes in CSI, CSI is utilized as the primary sensing signal since it implicitly encodes the information of target’s movement. In practice, raw CSI is strongly corrupted by amplitude and phase impairments. Therefore, noise reduction is first conducted before Doppler extraction. Subsequently, multi-link DFS are fused via PF, in which the target state is sequentially propagated and updated using likelihoods derived from Doppler measurements. This formulation naturally facilitates dynamic link selection and probabilistic fusion, as unreliable links are down-weighted through their likelihood contributions rather than being deterministically discarded. Moreover, the inaccuracy of device locations can be absorbed into measurement noise, contributing to the robustness against device-location uncertainty. In the absence of prior knowledge of device locations, self-localization is enabled first by jointly estimating the AoA and ToF of the Line of Sight (LoS) path between the AP and each STA. Preliminary self-localization tests of devices (Fig. 5) demonstrate a median STA positioning error of approximately 0.56 m (Table 1), which provides a relatively reliable initialization for the tracking algorithm. Results and Discussions A prototype is implemented for evaluation under a multi-transmitter-single-receiver architecture mounted with commodity Intel AX200/AX201 WiFi cards. CSI is collected in a 20 MHz bandwidth channel at a 5.24 GHz center frequency using PicoScenes, with 802.11ac packets transmitted at 100 to 200 Hz. Each CSI sample contains measurements of 57 subcarriers. In parallel, FTM measurement is performed at 2.4 GHz in channel 11 using iw and hostapd. In the prototype, each STA transmitter is equipped with a single antenna, while the AP receiver is equipped with two antennas. Experiments are conducted with one AP and three to four STAs (Fig. 7). WiTraj and PITrack using DR are set as baselines. When device self-localization is required, the proposed method achieves a median tracking error of 0.47 m, compared to 1.78 m for WiTraj and 1.77 m for PITrack (Fig. 9). This demonstrates an accuracy improvement of approximately 70% over traditional DR-based methods. Error injection experiments involving random positional disturbances reveal that, unlike traditional DR-based methods which are highly sensitive to location mismatch, the proposed approach exhibits robustness to device-location uncertainty (Fig. 10). Additionally, computational complexity analyses indicate that while execution time increases with the number of particles (Table 3), tracking performance stabilizes beyond a certain threshold (Fig. 11). Thus, real-time operation is attainable without sacrificing tracking quality by selecting an optimal particle count. Finally, evaluations in complex environments confirm that the proposed method consistently maintains accurate tracking performance, which is better than the baselines (Fig. 12). Conclusions A Particle-Filter-based approach is proposed for multi-link collaborative passive human tracking using commodity WiFi. By probabilistically fusing multiple links, robust tracking is achieved even when device locations contain errors. When device locations are unavailable, self-localization is supported through combining CSI-based estimation and FTM measurements, which supplies the necessary geometric priors for tracking. A prototype is built on a multi-transmitter-single-receiver network structure, which facilitates data aggregation and exploits AP-side multiple-antenna advantage. Experimental results indicate that (1) a sub-meter level tracking accuracy can be achieved when there exist device-location errors, which is much higher than the traditional method; (2) the proposed method has certain robustness to device-location uncertainty; and (3) the computational cost can meet real-time requirements when the particle count is properly chosen. However, the current study focuses on single-person tracking only. Future work should extend the framework to multi-person scenarios and evaluate performance under more diverse and realistic environmental conditions.
- WiFi sensing /
- Multi-link /
- Passive human tracking /
- Doppler Frequency Shift (DFS)

HTML全文

图 1 智能家居中的WiFi多链路跟踪系统模型

下载: 全尺寸图片幻灯片

图 2 系统流程图

下载: 全尺寸图片幻灯片

图 3 设备自定位示意图

下载: 全尺寸图片幻灯片

图 4 FTM协议原理示意图

下载: 全尺寸图片幻灯片

图 5 设备自定位测试场景示意图

下载: 全尺寸图片幻灯片

图 6 多普勒提取过程示意图

下载: 全尺寸图片幻灯片

图 7 实验场景示意图

下载: 全尺寸图片幻灯片

图 8 部分跟踪结果示意图

下载: 全尺寸图片幻灯片

图 9 跟踪误差的累积分布函数曲线

下载: 全尺寸图片幻灯片

图 10 设备位置误差对跟踪性能的影响

下载: 全尺寸图片幻灯片

图 11 不同参数下的跟踪性能

下载: 全尺寸图片幻灯片

图 12 复杂环境测试场景与结果示意图

下载: 全尺寸图片幻灯片

表 1 设备自定位测试结果

位置编号	1	2	3	4	5	6	7	8	9	10	11	12	13	14
AoA估计误差 (°)	6.2	8.2	5.0	11.6	13.4	0.0	3.0	10.4	2.0	16.0	22.6	5.4	1.8	5.4
FTM测距误差 (m)	0.07	0.00	0.35	0.17	0.11	0.21	0.08	0.33	0.54	0.42	0.29	0.82	0.68	0.59
STA定位误差 (m)	0.33	0.36	0.42	0.65	0.56	0.21	0.14	0.56	0.54	0.63	1.00	0.84	0.68	0.62

下载: 导出CSV

1 基于粒子滤波的多链路跟踪

输入：每条链路$ n $的DFS序列$ \left\{\overrightarrow{f_{\text{D}}^{n}}\right\} $，每个STA$ n $的位置$ \left(x_{\text{STA}}^{n},y_{\text{STA}}^{n}\right) $，链路数$ N $，粒子数$ {N}_{\text{S}} $，初始状态分布$ \boldsymbol{P} $，过程噪声协方差矩阵　$ \boldsymbol{Q} $，测量噪声方差$ {\sigma }^{2} $，时间间隔$ \Delta T $
输出：目标的估计状态$ {\left\{{\hat{\boldsymbol{x}}}[t]\right\}}_{t=1,2,\cdots ,T} $
1 for $ i $= 1 to$ {N}_{\text{S}} $ do
2 从初始状态分布中采样第$ i $个粒子：$ {\boldsymbol{x}}_{i}[1]\sim \boldsymbol{P} $；设置第$ i $个粒子的初始权重：$ {w}_{i}[1]=1/{N}_{\text{S}} $
3 end for
4 获取粒子初始状态：$ {\hat{\boldsymbol{x}}}[1]=\displaystyle\sum \limits_{i=1}^{{N}_{\text{S}}}{w}_{i}[1]{\boldsymbol{x}}_{i}[1] $
5 for $ t $= 2 to$ T $ do
6 for $ i $= 1 to$ {N}_{\text{S}} $ do
7 通过式(15)传播粒子$ {\boldsymbol{x}}_{i}[t] $
8 for $ n $= 1 to$ N $ do
9 通过式估计第$ i $个粒子的DFS$ \hat{f}_{\text{D}i}^{n}[t] $；通过式计算第$ i $个粒子的似然度$ {p}_{i} $；计算权重$ {w}_{i}[t]={w}_{i}[t]\times {p}_{i} $
10 end for
11 end for
12 权重归一化：$ {w}_{i}[t]={w}_{i}[t]/\displaystyle\sum \limits_{i=1}^{{N}_{\text{S}}}{w}_{i}[t] $；估计目标状态：$ {\hat{\boldsymbol{x}}}[t]=\displaystyle\sum \limits_{i=1}^{{N}_{\text{S}}}{w}_{i}[t]{\boldsymbol{x}}_{i}[t] $
13 重采样$ {N}_{\text{S}} $个粒子
14 for $ i $= 1 to$ {N}_{\text{S}} $ do
15 重置权重：$ {w}_{i}[t]=1/{N}_{\text{S}} $
16 end for
17 end for
18 return $ {\left\{{\hat{\boldsymbol{x}}}[t]\right\}}_{t=1,2,\cdots ,T} $

下载: 导出CSV

表 2 基于粒子滤波的多链路跟踪算法参数设置

$ {N}_{\text{S}} $	$ \boldsymbol{P} $	$ \boldsymbol{Q} $	$ {\sigma }^{2} $	$ \Delta T $
1000	$ \mathcal{N}\left(\left[\begin{array}{c}{x}_{\text{tar}}[1]\\{y}_{\text{tar}}[1]\\0\\0\end{array}\right],\text{diag}\left[\begin{array}{c}{0}^{2}\\{0}^{2}\\{0.01}^{2}\\{0.01}^{2}\end{array}\right]\right) $	$ \mathcal{N}\left(\left[\begin{array}{c}0\\0\\0\\0\end{array}\right],\text{diag}\left[\begin{array}{c}{0.05}^{2}\\{0.05}^{2}\\{0.25}^{2}\\{0.25}^{2}\end{array}\right]\right) $	0.4²	0.1 s

下载: 导出CSV

表 3 不同参数下的算法复杂度

参数	N	3				4
参数	N_S	500	1000	2000	5000	500	1000	2000	5000
运行时间 (s)	平均值	0.0138	0.0275	0.0551	0.1465	0.0137	0.0270	0.0553	0.1521
运行时间 (s)	最大值	0.0454	0.0808	0.1061	0.2952	0.0440	0.0654	0.1110	0.5602

下载: 导出CSV

参考文献(21)

[1]	LU Shihang, LIU Fan, LI Yunxin, et al. Integrated sensing and communications: Recent advances and ten open challenges[J]. IEEE Internet of Things Journal, 2024, 11(11): 19094–19120. doi: 10.1109/JIOT.2024.3361173.
[2]	ZHANG Tingting, MOLISCH A F, SHEN Yuan, et al. Joint power and bandwidth allocation in wireless cooperative localization networks[J]. IEEE Transactions on Wireless Communications, 2016, 15(10): 6527–6540. doi: 10.1109/TWC.2016.2580504.
[3]	SHI Fangzhan, LI Wenda, TANG Chong, et al. ML-Track: Passive human tracking using WiFi multi-link round-trip CSI and particle filter[J]. IEEE Transactions on Mobile Computing, 2025, 24(6): 5155–5172. doi: 10.1109/TMC.2025.3529897.
[4]	赵万龙, 田新元, 陈超, 等. 联合Spline插值的Wi-Fi指纹匹配定位算法[J]. 电子与信息学报, 2024, 46(9): 3563–3570. doi: 10.11999/JEIT230116. ZHAO Wanlong, TIAN Xinyuan, CHEN Chao, et al. Wi-Fi fingerprint localization uniting spline interpolation[J]. Journal of Electronics & Information Technology, 2024, 46(9): 3563–3570. doi: 10.11999/JEIT230116.
[5]	杨小龙, 张冰睿, 周牧, 等. 面向通感一体化的三维矩阵束联合参数估计算法[J]. 电子与信息学报, 2025, 47(1): 84–92. doi: 10.11999/JEIT240003. YANG Xiaolong, ZHANG Bingrui, ZHOU Mu, et al. A joint parameter estimation method based on 3D matrix pencil for integration of sensing and communication[J]. Journal of Electronics & Information Technology, 2025, 47(1): 84–92. doi: 10.11999/JEIT240003.
[6]	QIAN Kun, WU Chenshu, ZHANG Yi, et al. Widar2.0: Passive human tracking with a single Wi-Fi link[C]. Proceedings of the 16th Annual International Conference on Mobile Systems, Applications, and Services, Munich, Germany, 2018: 350–361. doi: 10.1145/3210240.3210314.
[7]	TONG Xinyu, GE Weiping, TIAN Yichen, et al. NNE-Tracking: A neural network enhanced framework for device-free Wi-Fi tracking[J]. IEEE Transactions on Mobile Computing, 2024, 23(9): 8981–8998. doi: 10.1109/TMC.2024.3357569.
[8]	LI Wenwei, GAO Ruiyang, XIONG Jie, et al. WiFi-CSI difference paradigm: Achieving efficient Doppler speed estimation for passive tracking[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2024, 8(2): 63. doi: 10.1145/3659608.
[9]	WU Dan, ZENG Youwei, GAO Ruiyang, et al. WiTraj: Robust indoor motion tracking with WiFi signals[J]. IEEE Transactions on Mobile Computing, 2023, 22(5): 3062–3078. doi: 10.1109/TMC.2021.3133114.
[10]	NIU Kai, WANG Xuanzhi, ZHANG Fusang, et al. Rethinking Doppler effect for accurate velocity estimation with commodity WiFi devices[J]. IEEE Journal on Selected Areas in Communications, 2022, 40(7): 2164–2178. doi: 10.1109/JSAC.2022.3155523.
[11]	XIE Yaxiong, XIONG Jie, LI Mo, et al. mD-Track: Leveraging multi-dimensionality for passive indoor Wi-Fi tracking[C]//The 25th Annual International Conference on Mobile Computing and Networking, Los Cabos, Mexico, 2019: 8. doi: 10.1145/3300061.3300133.
[12]	YAO Junmei, LIU Chaoyang, LUO Sheng, et al. Toward integrated sensing and communication: Interference-resistance design for WiFi sensing[J]. IEEE Transactions on Mobile Computing, 2025, 24(10): 10807–10822. doi: 10.1109/TMC.2025.3570752.
[13]	ZENG Youwei, WU Dan, XIONG Jie, et al. FarSense: Pushing the range limit of WiFi-based respiration sensing with CSI ratio of two antennas[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2019, 3(3): 121. doi: 10.1145/3351279.
[14]	杨小龙, 张亭亭, 周牧, 等. 面向通感一体化的变分模态分解-希尔伯特-黄变换呼吸频率感知算法[J]. 电子与信息学报, 2025, 47(4): 1014–1025. doi: 10.11999/JEIT240640. YANG Xiaolong, ZHANG Tingting, ZHOU Mu, et al. Variational mode decomposition-Hilbert-Huang transform breathing rate sensing algorithm for integration of sensing and communication[J]. Journal of Electronics & Information Technology, 2025, 47(4): 1014–1025. doi: 10.11999/JEIT240640.
[15]	KOTARU M, JOSHI K, BHARADIA D, et al. SpotFi: Decimeter level localization using WiFi[C]//Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication, London, UK, 2015: 269–282. doi: 10.1145/2785956.2787487.
[16]	ELEFTHERAKIS S, SANTAROMITA G, REA M, et al. SPRING+: Smartphone positioning from a single WiFi access point[J]. IEEE Transactions on Mobile Computing, 2024, 23(10): 9549–9566. doi: 10.1109/TMC.2024.3367241.
[17]	RATNAM V V, CHEN Hao, CHANG H H, et al. Optimal preprocessing of WiFi CSI for sensing applications[J]. IEEE Transactions on Wireless Communications, 2024, 23(9): 10820–10833. doi: 10.1109/TWC.2024.3376332.
[18]	WANG Zhongqin, ZHANG J A, ZHANG Haimin, et al. Passive human tracking with WiFi point clouds[J]. IEEE Internet of Things Journal, 2025, 12(5): 5528–5543. doi: 10.1109/JIOT.2024.3487193.
[19]	KIM Y and LING Hao. Human activity classification based on micro-Doppler signatures using a support vector machine[J]. IEEE Transactions on Geoscience and Remote Sensing, 2009, 47(5): 1328–1337. doi: 10.1109/TGRS.2009.2012849.
[20]	ELFRING J, TORTA E, and VAN DE MOLENGRAFT R. Particle filters: A hands-on tutorial[J]. Sensors, 2021, 21(2): 438. doi: 10.3390/s21020438.
[21]	JIANG Zhiping, LUAN T H, REN Xincheng, et al. Eliminating the barriers: Demystifying Wi-Fi baseband design and introducing the PicoScenes Wi-Fi sensing platform[J]. IEEE Internet of Things Journal, 2022, 9(6): 4476–4496. doi: 10.1109/JIOT.2021.3104666.