非接觸式生命徵象監測在長照／護理機構與居家場域備受關注，這些環境需要兼具連續性、低負擔與隱私的生理量測。此類場域常為小規模同處空間（同時在場人數≤6），而呼吸作為最即時、對惡化最敏感的指標之一，對慢性呼吸疾病（如 COPD）、睡眠呼吸障礙與感染性疾病（如 COVID-19）的早期偵測與風險評估至關重要。傳統接觸式感測存在舒適度與活動受限等問題；影像式方法又受光照、隱私與感染風險限制，因此雷達被視為具潛力的替代方案。然而，在多人情境中受測人數判定相當困難：當多位個體距離相近或呼吸頻率相近時，回波會嚴重重疊而使分離變得困難；在單天線系統中，由於空間解析度受限，僅依賴頻譜峰值或陣列波束形成的方法常常失效。本研究將二維子空間自動模型階數選擇（2D-SAMOS）整合至 FMCW 雷達處理流程，提出一個模型階數估計框架，無需頻譜分離或天線陣列。透過建構時間域 Hankel 矩陣並分析奇異值行為，所提方法能有效捕捉潛在訊號源的能量分布，即使在呼吸頻率極為接近的情況下，仍能穩定推斷同時受測人數。實驗結果顯示，即便在單天線前端與嚴重重疊條件下，仍具高準確度與穩健性。所提方法為護理病房與居家監測提供輕量、可攜的前端解決方案。

Non-contact vital-sign monitoring has attracted significant attention in long-term care (LTC)/nursing facilities and at-home monitoring, where continuous, low-burden, and privacy-preserving physiological measurement is increasingly required. These settings demand reliable respiration tracking in small-occupancy rooms (≤6 concurrent subjects). Among vital signs, respiration is one of the most immediate and sensitive indicators of deterioration, crucial for early detection and risk assessment in chronic respiratory diseases (e.g., COPD), sleep-disordered breathing, and infectious illnesses such as COVID-19. Traditional contact-based sensors suffer from issues of comfort and restricted mobility, while camera-based methods are constrained by illumination, privacy concerns, and infection risk—motivating radar sensing as a promising alternative. However, accurately determining the number of subjects remains challenging in multi-target environments. When multiple individuals are at similar ranges or exhibit closely spaced respiration rates, radar echoes substantially overlap, complicating source separation. This limitation is amplified in single-antenna systems with restricted spatial resolution, where approaches relying on spectral peaks or array beamforming often fail. This work integrates two-dimensional Subspace Automatic Model Order Selection (2D-SAMOS) into the FMCW radar processing chain, proposing a model-order selection framework that does not require spectral separation or antenna-array beamforming. By constructing timedomain Hankel matrices and analyzing singular-value behavior, the framework captures the energy distribution of latent sources and reliably infers the number of concurrent subjects even under closely spaced respiration frequencies. Experimental results demonstrate high accuracy and robustness in the presence of severe signal overlap while operating with a singleantenna setup. The proposed method offers a lightweight, portable front-end solution tailored to LTC/nursing rooms and home monitoring.

[1]M. Rehman, N. A. A. Ali, R. A. Shah, M. B. Khan, S. A. Shah, A. Alomainy, X. Yang, M. A. Imran, and Q. H. Abbasi, “Development of an intelligent real-time multiperson respiratory illnesses sensing system using SDR technology,” IEEE Sensors Journal, vol. 22, no. 19, pp. 18858–18869, 2022.
[2]J. Zhang, R. Xi, Y. He, Y. Sun, X. Guo, W. Wang, X. Na, Y. Liu, Z. Shi, and T. Gu, “A survey of mmWave-based human sensing: Technology, platforms and applications,” IEEE Communications Surveys & Tutorials, vol. 25, no. 4, pp. 2052–2087, 2023.
[3]T. Hussain, S. Ullah, R. Fernandez-Garcia, and I. Gil, “Wearable sensors for respiration monitoring: A review,” Biosensors, vol. 13, no. 9, p. 823, 2023.
[4]A. Singh, S. U. Rehman, S. Yongchareon, and P. H. J. Chong, “Multi-resident non-contact vital sign monitoring using radar: A review,” IEEE Sensors Journal, vol. 21, no. 4, pp. 4061–4084, 2021.
[5]F.-K. Wang, C.-T. M. Wu, T.-S. Horng, C.-H. Tseng, S.-H. Yu, C.-C. Chang, P.-H. Juan, and Y. Yuan, “Review of self-injection-locked radar systems for noncontact detection of vital signs,” IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology, vol. 4, no. 4, pp. 294–307, 2020.
[6]A. T. Purnomo, K. S. Komariah, D.-B. Lin, W. F. Hendria, B.-K. Sin, and N. Ahmadi, “Non-contact supervision of COVID-19 breathing behaviour with FMCW radar and stacked ensemble learning model in real-time,” IEEE Transactions on Biomedical Circuits and Systems, vol. 16, no. 4, pp. 664–678, 2022.
[7]Health Building Note 00-03: Clinical and Clinical Support Spaces, NHS England, 2014. [Online]. Available: https://www.england.nhs.uk/publication/designing-generic-clinical-and-clinical-support-spaces-hbn-00-03/
[8]Code of Federal Regulations, “42 CFR 483.90 physical environment,” 2024. Bedrooms must accommodate ≤4 residents; new facilities after Nov 28, 2016 must accommodate ≤2 residents. [Online]. Available: https://www.ecfr.gov/current/title-42/part-483/section-483.90
[9]J. Martos, E. Lane, S. Preston, S. E. F. Spencer, and T. House, “Modelling the impact of ward design on healthcare-associated infections: A case study of four-bed vs six-bed bays,” Mathematical Biosciences and Engineering, vol. 17, no. 6, pp. 6955–6974, 2020.
[10]Demographic Statistics Quarterly, Autumn 2024 (人口統計季刊113年秋季), Department of Household Registration, Ministry of the Interior, 2024. Summary indicates an average of 2.5 persons per household. [Online]. Available: https://www.ris.gov.tw/documents/data/5/2/DemographicQuarterly_Autumn2024.pdf
[11]M. S. Islam et al., “Radar-based remote physiological sensing: Progress, challenges, and opportunities,” Frontiers in Physiology, vol. 13, p. 955208, 2022.
[12]Y. Eder, Z. Liu, and Y. C. Eldar, “Sparse non-contact multiple people localization and vital signs monitoring via FMCW radar,” in Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023.
[13]C.-H. Hsieh and P.-H. Tseng, “Multiperson localization and vital signs estimation using mmWave MIMO radar,” IEEE Transactions on Microwave Theory and Techniques, vol. 73, no. 2, pp. 1222–1234, 2025.
[14]X. Li, Q. Zhang, J. Wang, W. Chen, and R. Zhang, “Multiperson detection and vital-sign sensing empowered by space-time-coding reconfigurable intelligent surfaces,” IEEE Internet of Things Journal, vol. 11, no. 17, pp. 28876–28890, 2024.
[15]X. Yang, H. Zhao, Z. Zhuang, D. Wang, H. Hong, and X. Zhu, “Super-range-resolution respiratory signals detection from multiple targets with single-channel FMCW radar,” IEEE Transactions on Instrumentation and Measurement, vol. 74, p. 4006015, 2025.
[16]C.-H. Lin, C.-Y. Chi, L. Chen, D. J. Miller, and Y. Wang, “Detection of sources in non-negative blind source separation by minimum description length criterion,” IEEE Transactions on Neural Networks and Learning Systems, vol. 29, no. 9, pp. 4022–4037, 2018.
[17]H. Akaike, “A new look at the statistical model identification,” IEEE Transactions on Automatic Control, vol. 19, no. 6, pp. 716–723, 1974.
[18]G. Schwarz, “Estimating the dimension of a model,” The Annals of Statistics, vol. 6, no. 2, pp. 461–464, 1978.
[19]N. Asadi and B. Seyfe, “Model order estimation based on the correntropy of observation eigenvalues,” Signal Processing, vol. 216, p. 109310, 2024.
[20]R. Badeau, G. Richard, and B. David, “Selecting the modeling order for the ESPRIT high resolution method: An alternative approach (ESTER),” in Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2004, pp. 1025–1028.
[21]M. Mercuri, G. Sacco, R. Hornung, P. Zhang, H. J. Visser, M. Hijdra, Y.-H. Liu, S. Pisa, B. van Liempd, and T. Torfs, “2-D localization, angular separation and vital signs monitoring using a SISO FMCW radar for smart long-term health monitoring environments,” IEEE Internet of Things Journal, vol. 8, no. 14, pp. 11065–11077, 2021.