基于深度盲源分离的毫米波雷达驾驶员心跳感知方法研究

丁凌娜; 张宾宾; 王健阳; 张东恒; 陈彦

doi:10.12000/JR26042

基于深度盲源分离的毫米波雷达驾驶员心跳感知方法研究

DOI: 10.12000/JR26042 CSTR: 32380.14.JR26042

中国科学技术大学网络空间安全学院合肥 230026

基金项目: 国家自然科学基金(62201542)，中央高校基本科研业务费专项资金资助

详细信息

作者简介:
丁凌娜，硕士，主要研究方向为无线感知

张宾宾，博士，主要研究方向为无线感知

王健阳，博士，主要研究方向为无线感知

张东恒，副研究员，主要研究方向为无线感知

陈　彦，教授，主要研究方向为多模态感知

通讯作者:
张东恒 dongheng@ustc.edu.cn

责任主编：方震 Corresponding Editor: FANG Zhen

中图分类号: TN959.6
计量
- 文章访问数:
- HTML全文浏览量:
- PDF下载量:
- 被引次数: 0
出版历程
- 收稿日期: 2026-02-09

Driver Heartbeat Sensing using Millimeter-Wave Radar based on Deep Blind Source Separation

School of Cyber Science and Technology, University of Science and Technology of China, Hefei 230026, China

Funds: The National Natural Science Foundation of China (62201542), The Fundamental Research Funds for the Central Universities

More Information

Corresponding author: ZHANG Dongheng, dongheng@ustc.edu.cn

摘要

摘要: 对驾驶员心脏活动的监测可以有效评估其生理和心理状态，然而，现有的心跳感知方法(心电图和远程光电容积描记法)操作繁琐，易受光线影响，并不适用于车载场景。基于毫米波雷达的心跳感知技术虽然具有高精度、非接触等优势，但易受干扰影响。针对上述问题，该文结合射频信号的低频特性、长程动态敏感性、稀疏性，基于自注意力机制设计了一个射频特征提取器，通过构建深度盲源分离网络实现了驾驶员心跳射频特征和车载干扰特征的分离。此外，针对射频信号采集难的问题，该文提出了一种混合-源信号生成策略，通过少量心震描记数据和车载干扰数据合成了训练样本。最后，该文在真实的行车环境中对该方法进行了验证，实验结果表明，系统可实现心率绝对误差4.92 bpm、心搏间期中位数误差65.93 ms的感知精度。
- 无线感知 /
- 心跳感知 /
- 盲源分离 /
- 深度学习 /
- 智能驾驶
Abstract: Monitoring of drivers’ cardiac activity enables effective assessment of their physiological and psychological states. However, existing methods such as electrocardiography and remote phontoplethysmograyhy are cumbersome and sensitive to lighting conditions, limiting their applicability in vehicular settings. Despite its high accuracy and noncontact operation, millimeter-wave radar-based heartbeat sensing is inherently vulnerable to interference. To address these issues, this paper exploits the low-frequency characteristics, long-range dynamic sensitivity, and sparsity of radio-frequency (RF) signals and designs a self-attention-based RF feature extractor. On this basis, a deep blind source separation network is constructed to separate the driver’s heartbeat-related RF features from in-vehicle interference. Furthermore, to reduce the burden of RF signal acquisition, we introduce a hybrid-source signal generation strategy that synthesizes a large number of mixed and ground-truth source signals using only a small number of seismocardiogram and interference signals. Finally, extensive on-road testing demonstrates that the proposed system achieves a median heart rate error of 4.92 bpm and a median interbeat interval error of 65.93 ms.
- Wireless sensing /
- Heartbeat sensing /
- Blind source separation /
- Deep learning /
- Intelligent driving

HTML全文

图 1 系统框图

Figure 1. System block diagram

下载: 全尺寸图片幻灯片

图 2 杂波消除算法效果图

Figure 2. Visualization of the clutter removal

下载: 全尺寸图片幻灯片

图 3 驾驶员定位算法示意图

Figure 3. Visualization of the driver localization

下载: 全尺寸图片幻灯片

图 4 呼吸干扰抑制示意图

Figure 4. Illustration of respiratory interference suppression

下载: 全尺寸图片幻灯片

图 5 心跳分离算法流程图

Figure 5. Illustration of the heartbeat separation algorithm

下载: 全尺寸图片幻灯片

图 6 特征掩码分离器的结构

Figure 6. Architecture of the feature-mask separator

下载: 全尺寸图片幻灯片

图 7 混合-源信号生成训练数据示意图

Figure 7. Illustration of hybrid-source signal generation strategy

下载: 全尺寸图片幻灯片

图 8 毫米波雷达设备

Figure 8. Millimeter-wave radar device

下载: 全尺寸图片幻灯片

图 9 心电传感器

Figure 9. Electrocardiogram sensor

下载: 全尺寸图片幻灯片

图 10 测试数据采集场景

Figure 10. Experimental setup

下载: 全尺寸图片幻灯片

图 11 模型训练情况

Figure 11. Model training results

下载: 全尺寸图片幻灯片

图 12 模型测试情况

Figure 12. Model testing result

下载: 全尺寸图片幻灯片

图 13 心率估计误差累积分布函数

Figure 13. Cumulative distribution function of heart rate estimation errors

下载: 全尺寸图片幻灯片

图 14 心搏间期估计误差累积分布函数

Figure 14. Cumulative distribution function of IBI estimation errors

下载: 全尺寸图片幻灯片

表 1 雷达参数

Table 1. Radar parameters

雷达配置	具体参数
起始频率$ {\mathrm{f}}_{\mathrm{c}} $	60 GHz
带宽$ \mathrm{B} $	3.9 GHz
啁啾持续时间$ {\mathrm{T}}_{\mathrm{c}} $	60 ms
距离单元$ \mathrm{R} $	30
相位信号采样率$ {\mathrm{f}}_{\mathrm{s}} $	100 Hz

下载: 导出CSV

表 2 心率估计误差统计结果

Table 2. Heart rate estimation error statistics

方案	HR_RMSE	HR_MAE	HR_MED
基线方案1	10.46bpm	9.49bpm	10.12bpm
基线方案2	5.26bpm	4.35bpm	4.17bpm
本文方案	3.79 bpm	3.35 bpm	3.16 bpm

下载: 导出CSV

表 3 心搏间期估计误差统计结果

Table 3. IBI estimation error statistics

方案	IBI_RMSE	IBI_MAE	IBI_MED
基线方案1	148.02ms	125.94ms	120.00ms
基线方案2	97.16 ms	78.42 ms	67.29 ms
本文方案	45.13 ms	41.87 ms	39.23 ms

下载: 导出CSV

表 4 不同模型结构、训练方式下心率、心搏间期估计误差结果

Table 4. HR and IBI estimation errors under different model structures and training strategies

特征提取器	损失函数	叠加方式	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
本文结构	SI-SNR	线性叠加	3.79 bpm	3.35 bpm	3.16 bpm	45.13 ms	41.87 ms	39.23 ms
	SI-SNR	非线性叠加	4.23 bpm	3.53 bpm	3.22 bpm	47.70 ms	46.90 ms	42.79 ms
	Cos-Sim	线性叠加	4.36 bpm	3.61 bpm	3.27 bpm	48.99 ms	46.56 ms	42.93 ms
	Cos-Sim	非线性叠加	4.52 bpm	3.78 bpm	3.30 bpm	49.11 ms	46.79 ms	43.42 ms
mmformer	SI-SNR	线性叠加	4.60 bpm	3.81 bpm	3.34 bpm	49.33 ms	46.93 ms	44.18 ms
	SI-SNR	非线性叠加	4.99 bpm	4.23 bpm	3.76 bpm	51.32 ms	49.04 ms	46.14 ms
	Cos-Sim	线性叠加	4.64 bpm	3.83 bpm	3.40 bpm	50.50 ms	48.50 ms	45.22 ms
	Cos-Sim	非线性叠加	5.03 bpm	4.25 bpm	3.71 bpm	52.81 ms	50.18 ms	46.47 ms
Convtasnet	SI-SNR	线性叠加	5.04 bpm	4.30 bpm	3.80 bpm	55.03 ms	52.22 ms	48.12 ms
	SI-SNR	非线性叠加	6.23 bpm	5.36 bpm	4.79 bpm	59.13 ms	54.43 ms	52.15 ms
	Cos-Sim	线性叠加	5.62 bpm	5.32 bpm	4.75 bpm	57.50 ms	52.58 ms	48.46 ms
		非线性叠加	6.68 bpm	5.89 bpm	4.87 bpm	67.50 ms	62.58 ms	53.27 ms

下载: 导出CSV

表 5 不同训练数据集下心率、心搏间期估计误差结果

Table 5. HR and IBI estimation errors under different training datasets

	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
5000对	7.18bpm	6.39bpm	5.51bpm	87.53ms	74.11 ms	60.00ms
10000对	5.96bpm	5.53bpm	5.05bpm	63.67 ms	62.16 ms	59.47ms
25000对	4.77bpm	3.86bpm	3.72 bpm	54.36ms	49.23 ms	47.87ms
50000对	3.79 bpm	3.35 bpm	3.16 bpm	45.13 ms	41.87 ms	39.23 ms

下载: 导出CSV

表 6 不同受试者的心率、心搏间期估计误差结果

Table 6. HR and IBI estimation errors across different subjects

	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
受试者1	4.52 bpm	3.92 bpm	3.07 bpm	44.00ms	43.28 ms	41.24 ms
受试者2	3.42 bpm	2.99 bpm	3.19 bpm	44.65 ms	43.82 ms	44.71 ms
受试者3	4.01 bpm	3.84 bpm	3.05 bpm	40.12 ms	38.72 ms	38.27 ms
受试者4	2.50 bpm	2.26 bpm	2.17 bpm	35.95 ms	34.31 ms	31.43 ms
受试者5	4.09 bpm	3.58 bpm	2.88 bpm	59.54 ms	53.85 ms	50.57 ms
受试者6	3.32 bpm	3.31 bpm	3.38 bpm	53.31 ms	50.37 ms	45.06 ms
受试者7	4.06 bpm	3.21 bpm	2.07 bpm	43.84 ms	41.02 ms	36.76 ms
受试者8	4.02 bpm	3.71 bpm	3.23 bpm	40.77 ms	39.60 ms	40.45 ms

下载: 导出CSV

表 7 不同车型条件下心率、心搏间期估计误差结果

Table 7. HR and IBI estimation errors under different vehicle types

	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
油车	4.06 bpm	3.21 bpm	2.07 bpm	43.84 ms	41.02 ms	36.76 ms
电车	3.37 bpm	3.05 bpm	2.97 bpm	33.16 ms	31.32 ms	28.47 ms

下载: 导出CSV

表 8 不同路面情况下心率、心搏间期估计误差结果

Table 8. HR and IBI estimation errors under different road conditions

	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
良好	3.42 bpm	2.99 bpm	3.19 bpm	44.65 ms	43.82 ms	44.71 ms
一般	5.13 bpm	5.09 bpm	4.94 bpm	87.25 ms	62.29 ms	61.70 ms
差	7.64 bpm	6.67 bpm	6.19 bpm	101.08 ms	94.65 ms	91.39 ms
综合	5.40 bpm	4.92 bpm	4.77 bpm	77.66 ms	66.92 ms	65.93 ms

下载: 导出CSV

表 9 不同驾驶环境下心率、心搏间期估计误差结果

Table 9. HR and IBI estimation errors under under different driving environments

	HR_RMSE	HR_MAE	HR_MED	IBI_RMSE	IBI_MAE	IBI_MED
静态(未启动)	1.28 bpm	1.17bpm	0.99bpm	30.87ms	30.73ms	29.31ms
静态(已启动)	1.90 bpm	1.59bpm	1.51bpm	47.31ms	43.36ms	40.00ms
动态	3.32 bpm	3.31 bpm	3.38 bpm	53.31 ms	50.37 ms	45.06 ms

下载: 导出CSV

参考文献(48)

[1]	World Health Organization. World health statistics 2025[R/OL]. Geneva: World Health Organization, 2025[2026-05-12]. https://www.who.int/publications/b/78420.
[2]	国家统计局. 中国统计年鉴2024:24-5 道路交通事故情况(2023年)[M/OL]. 北京: 中国统计出版社, 2024[2026-05-12]. https://www.stats.gov.cn/sj/ndsj/2024/indexch.htm. National Bureau of Statistics of China. China Statistical Yearbook 2024: 24-5 Traffic Accidents (2023)[M/OL]. Beijing: China Statistics Press,2024[2026-05-12].https://www.stats.gov.cn/sj/ndsj/2024/indexch.htm.
[3]	DU Guanglong, ZHANG Linlin, SU Kang, et al. A multimodal fusion fatigue driving detection method based on heart rate and PERCLOS[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(11): 21810–21820. doi: 10.1109/TITS.2022.3176973.
[4]	POP G N, CHRISTODORESCU R, VELIMIROVICI D E, et al. Assessment of the impact of alcohol consumption patterns on heart rate variability by machine learning in healthy young adults[J]. Medicina, 2021, 57(9): 956. doi: 10.3390/medicina57090956.
[5]	HASNUL M A, AZIZ N A A, ALELYANI S, et al. Electrocardiogram-based emotion recognition systems and their applications in healthcare—a review[J]. Sensors, 2021, 21(15): 5015. doi: 10.3390/s21155015.
[6]	CHEN Weixuan and MCDUFF D. Deepphys: Video-based physiological measurement using convolutional attention networks[C]. 15th European Conference on Computer Vision – ECCV 2018, Munich, Germany, 2018: 356–373. doi: 10.1007/978-3-030-01216-8_22.
[7]	YU Zitong, SHEN Yuming, SHI Jingang, et al. PhysFormer: Facial video-based physiological measurement with temporal difference transformer[C]. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA, 2022: 4176–4186. doi: 10.1109/CVPR52688.2022.00415.
[8]	ZOU Bochao, GUO Zizheng, CHEN Jiansheng, et al. RhythmFormer: Extracting patterned rPPG signals based on periodic sparse attention[J]. Pattern Recognition, 2025, 164: 111511. doi: 10.1016/j.patcog.2025.111511.
[9]	CHOI J and LEE S J. MMDrive: Multi-modal remote physiological signal measurement dataset for driver status monitoring[C]. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Nashville, USA, 2025: 5691–5698. doi: 10.1109/CVPRW67362.2025.00567.
[10]	DAI Tianyang, CHEN Yan, and HU Yang. rPPG-NDCL: Unsupervised remote physiological measurement via noise-disentangled contrastive learning[C]. 2025 IEEE International Conference on Image Processing (ICIP), Anchorage, USA, 2025: 2856–2861. doi: 10.1109/ICIP55913.2025.11084290.
[11]	LIANG Qian, CHEN Yan, and Hu Yang. Continual learning for remote physiological measurement: Minimize forgetting and simplify inference[C]. 18th European Conference on Computer Vision – ECCV 2024, Milan, Italy, 2024: 126–144. doi: 10.1007/978-3-031-72764-1_8.
[12]	杨轩, 王子颖, 张力, 等. 基于盲源分离的多人呼吸信号检测方法[J]. 雷达学报(中英文), 2025, 14(1): 117–134. doi: 10.12000/JR24115. YANG Xuan, WANG Ziying, ZHANG Li, et al. Noncontact multiperson respiratory detection method based on blind source separation[J]. Journal of Radars, 2025, 14(1): 117–134. doi: 10.12000/JR24115.
[13]	赵翔, 王威, 李晨洋, 等. 基于毫米波雷达微动信号和脉搏波数据融合的睡眠呼吸暂停低通气综合征筛查技术[J]. 雷达学报(中英文), 2025, 14(1): 102–116. doi: 10.12000/JR24107. ZHAO Xiang, WANG Wei, LI Chenyang, et al. Diagnosis of sleep apnea hypopnea syndrome using fusion of micro-motion signals from millimeter-wave radar and pulse wave data[J]. Journal of Radars, 2025, 14(1): 102–116. doi: 10.12000/JR24107.
[14]	方震, 简璞, 张浩, 等. 基于FMCW雷达的非接触式医疗健康监测技术综述[J]. 雷达学报, 2022, 11(3): 499–516. doi: 10.12000/JR22019. FANG Zhen, JIAN Pu, ZHANG Hao, et al. Review of noncontact medical and health monitoring technologies based on FMCW radar[J]. Journal of Radars, 2022, 11(3): 499–516. doi: 10.12000/JR22019.
[15]	郑学召, 丁文, 黄渊, 等. 不同领域下超宽带雷达探测呼吸心跳信号研究综述[J]. 雷达学报, 2024, 14(1): 204–228. doi: 10.12000/JR24154. ZHENG Xuezhao, DING Wen, HUANG Yuan, et al. A review of UWB radar detection of respiration and heartbeat signals in different scenarios[J]. Journal of Radars, 2025, 14(1): 204–228. doi: 10.12000/JR24154.
[16]	李志, 唐成垚, 戴永鹏, 等. 多旋翼无人机载4D成像雷达生命体征感知方法[J]. 雷达学报(中英文), 2025, 14(1): 62–72. doi: 10.12000/JR24128. LI Zhi, TANG Chengyao, DAI Yongpeng, et al. Multirotor UAV-borne vital signs sensing using 4D imaging radar[J]. Journal of Radars, 2025, 14(1): 62–72. doi: 10.12000/JR24128.
[17]	刘畅宇, 张浩, 耿芳琳, 等. 基于距离抽头重构的生理雷达动态解调算法[J]. 雷达学报(中英文), 2025, 14(1): 135–150. doi: 10.12000/JR24151. LIU Changyu, ZHANG Hao, GENG Fanglin, et al. Dynamic demodulation algorithm for bio-radar sensors based on range tapper[J]. Journal of Radars, 2025, 14(1): 135–150. doi: 10.12000/JR24151.
[18]	GUO Zehan, ZHANG Binbin, CHEN Jinbo, et al. Contactless fine-grained cardiac events detection and segmentation with radio frequency signals[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2025, 9(3): 82. doi: 10.1145/3749457.
[19]	ZHANG Binbin, ZHANG Dongheng, LI Yadong, et al. Monitoring long-term cardiac activity with contactless radio frequency signals[J]. Nature Communications, 2024, 15(1): 10598. doi: 10.1038/s41467-024-55061-9.
[20]	YUAN Yuqin, CHEN Jinbo, ZHANG Dongheng, et al. Atrial fibrillation detection via contactless radio monitoring and knowledge transfer[J]. Nature Communications, 2025, 16(1): 4317. doi: 10.1038/s41467-025-59482-y.
[21]	GONG Hanqin, CHEN Jinbo, XU Guixin, et al. Osense: Omni-directional heartbeat sensing with radio signal[J]. IEEE Internet of Things Journal, 2025, 12(12): 25269–25285. doi: 10.1109/JIOT.2025.3558733.
[22]	XU Xiaohan, ZHANG Dongheng, LU Zhi, et al. Contactless nighttime stress monitoring with mmwave radar[C]. ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Hyderabad, India, 2025: 1–5. doi: 10.1109/ICASSP49660.2025.10890411.
[23]	ZHANG Haoyu, ZHANG Dongheng, SONG Ruiyuan, et al. UMIMO: Universal unsupervised learning for mmwave radar sensing with MIMO array synthesis[J]. IEEE Transactions on Mobile Computing, 2025, 24(8): 7042–7058. doi: 10.1109/TMC.2025.3546757.
[24]	LIN Jiayan, GUO Shisheng, XUE Shucheng, et al. Non-line-of-sight human vital sign detection method based on correntropy and cross-correlation with multipath exploitation[C]. 2025 IEEE Radar Conference (RadarConf25), Krakow, Poland, 2025: 1028–1033. doi: 10.1109/RadarConf2559087.2025.11205110.
[25]	GU Boyuan, YANG Yanhui, YOU Siyu, et al. Improved VMD based remote heartbeat estimation utilizing 60GHz mmWave radar[C]. 2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Vienna, Austria, 2025: 4635–4640. doi: 10.1109/SMC58881.2025.11342664.
[26]	LI Jiachen, GUO Shisheng, CUI Guolong, et al. Multidomain separation for human vital signs detection with FMCW radar in interference environment[J]. IEEE Transactions on Microwave Theory and Techniques, 2024, 72(7): 4278–4293. doi: 10.1109/TMTT.2023.3337101.
[27]	LIU Xingyao, CHEN Jiahui, GUO Shisheng, et al. Non-line-of-sight human vital sign detection method based on multipath signals fusion[C]. 2023 3rd International Conference on Frontiers of Electronics, Information and Computation Technologies (ICFEICT), Yangzhou, China, 2023: 489–496. doi: 10.1109/ICFEICT59519.2023.00086.
[28]	HA U, MADANI S, and ADIB F. WiStress: Contactless stress monitoring using wireless signals[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2021, 5(3): 103. doi: 10.1145/3478121.
[29]	LIANG Kun, ZHOU Anfu, ZHANG Zhan, et al. mmStress: Distilling human stress from daily activities via contact-less millimeter-wave sensing[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2023, 7(3): 110. doi: 10.1145/3610926.
[30]	ZHU Juncen, CAO Jiannong, YANG Yanni, et al. mmdrive: Fine-grained fatigue driving detection using mmwave radar[J]. ACM Transactions on Internet of Things, 2023, 4(4): 26. doi: 10.1145/3614437.
[31]	ZHENG Tianyue, CHEN Zhe, CAI Chao, et al. V2iFi: In-vehicle vital sign monitoring via compact RF sensing[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2020, 4(2): 70. doi: 10.1145/3397321.
[32]	XU Xiangyu, YU Jiadi, MA Chengguang, et al. mmECG: Monitoring human cardiac cycle in driving environments leveraging millimeter wave[C]. IEEE INFOCOM 2022 - IEEE Conference on Computer Communications, London, United Kingdom, 2022: 90–99. doi: 10.1109/INFOCOM48880.2022.9796912.
[33]	WANG Fengyu, ZENG Xiaolu, WU Chenshu, et al. Driver vital signs monitoring using millimeter wave radio[J]. IEEE Internet of Things Journal, 2022, 9(13): 11283–11298. doi: 10.1109/JIOT.2021.3128548.
[34]	HU Qingyong, ZHANG Qian, LU Hao, et al. Contactless arterial blood pressure waveform monitoring with mmwave radar[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2024, 8(4): 178. doi: 10.1145/3699781.
[35]	KHAN U M, RIGAZIO L, and SHAHZAD M. Contactless monitoring of PPG using radar[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2022, 6(3): 123. doi: 10.1145/3550330.
[36]	CHEN Jinbo, ZHANG Dongheng, WU Zhi, et al. Contactless electrocardiogram monitoring with millimeter wave radar[J]. IEEE Transactions on Mobile Computing, 2024, 23(1): 270–285. doi: 10.1109/TMC.2022.3214721.
[37]	WANG Haoyu, CHEN Jinbo, ZHANG Dongheng, et al. Contactless radar heart rate variability monitoring via deep spatio-temporal modeling[C]. ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Seoul, Korea, Republic of, 2024: 111–115. doi: 10.1109/ICASSP48485.2024.10447570.
[38]	Wang Jiyao, YANG Xiao, HU Qingyong, et al. Physdrive: A multimodal remote physiological measurement dataset for in-vehicle driver monitoring[J]. Advances in Neural Information Processing Systems, 2026, 38.
[39]	ZHANG Dongheng, HU Yang, and CHEN Yan. MTrack: Tracking multiperson moving trajectories and vital signs with radio signals[J]. IEEE Internet of Things Journal, 2021, 8(5): 3904–3914. doi: 10.1109/JIOT.2020.3025820.
[40]	ZHAO Mingmin, ADIB F, and KATABI D. Emotion recognition using wireless signals[C]. The 22nd Annual International Conference on Mobile Computing and Networking, New York, USA, 2016: 95–108. doi: 10.1145/2973750.2973762.
[41]	ÖZER Y and MÜLLER M. Source separation of piano concertos using musically motivated augmentation techniques[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2024, 32: 1214–1225. doi: 10.1109/TASLP.2024.3356980.
[42]	LUO Yi and MESGARANI N. Conv-tasnet: Surpassing ideal time–frequency magnitude masking for speech separation[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2019, 27(8): 1256–1266. doi: 10.1109/TASLP.2019.2915167.
[43]	STOLLER D, EWERT S, and DIXON S. Wave-U-Net: A multi-scale neural network for end-to-end audio source separation [C]. The 19th International Society for Music Information Retrieval Conference, Paris, France, 2018: 334–340.
[44]	DÉFOSSEZ A, USUNIER N, BOTTOU L, et al. Demucs: Deep extractor for music sources with extra unlabeled data remixed[J]. arXiv preprint arXiv:1909.01174, 2019.
[45]	HA U, ASSANA S, and ADIB F. Contactless seismocardiography via deep learning radars [C]. The 26th Annual International Conference on Mobile Computing and Networking, London, United Kingdom, 2020: 62. doi: 10.1145/3372224.3419982.
[46]	PARLATO S, CENTRACCHIO J, CINOTTI E, et al. A Forcecardiography dataset with simultaneous SCG, Heart Sounds, ECG, and Respiratory signals[J]. Scientific Data, 2025, 12(1): 1370. doi: 10.1038/s41597-025-05694-2.
[47]	CHEN Jiayu, SONG Yingjian, ZHANG Yida, et al. SelfDenoiser: Self-supervised seismic signal denoiser for continuous and contactless cardiac monitoring[J]. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2025, 9(4): 165. doi: 10.1145/3770701.
[48]	DROITCOUR A D, BORIC-LUBECKE O, LUBECKE V M, et al. Range correlation and I/Q performance benefits in single-chip silicon Doppler radars for noncontact cardiopulmonary monitoring[J]. IEEE Transactions on Microwave Theory and Techniques, 2004, 52(3): 838–848. doi: 10.1109/TMTT.2004.823552.