基于77 GHz FMCW雷达和小波散射变换的运动员运动前后生理皮肤状态监测
Pre- and Post-Exercise Monitoring of Physiological Skin States in Sportsmen Using 77 GHz FMCW Radar and Wavelet Scattering Transform
浏览论文内容
中文总结 AI 辅助
本文提出利用77 GHz FMCW雷达结合小波散射变换，非接触式分类运动员运动前后皮肤状态变化，在15人数据集上达到91.55%准确率，验证了边缘部署的可行性。
中文摘要 AI 辅助
本文提出了一种非接触式雷达传感框架，用于对固定时长剧烈体育活动及饮水限制后浅表皮肤层中生理性变化进行分类。采用77 GHz调频连续波（FMCW）雷达，在体育活动前后采集坐于雷达前方的受试者胸部高分辨率距离像。由于毫米波频率下电磁波对生物组织的穿透深度固有地较浅，传感主要局限于角质层和表皮上层。我们证明，运动引起的体温调节、汗液残留沉积以及浅表组织水合作用的瞬时变化会产生可测量的准静态雷达反射率变化。我们利用持续时间为102.4微秒的单线性调频快照来抑制如缓慢变化的心肺运动等混杂因素，随后通过小波散射变换（WST）进行轻量级特征提取，并使用适用于边缘节点的轻量级机器学习模型进行分类。对15名运动员在运动前后测量的数据集进行留一受试者交叉验证，平均分类准确率为91.55%，95%置信区间为[84.27, 98.83]%。可解释性分析显示，低阶散射系数（捕获总体信号能量和幅度稳定性）主导判别能力，与皮肤表面介电常数的准静态变化一致。这些结果为基于雷达、可边缘部署的运动后皮肤状态分类建立了概念验证，为运动科学中的非接触式生理监测开辟了新途径。
英文摘要
This paper presents a contactless radar sensing framework that classifies physiologically-induced changes in the superficial skin layer following intense fixed-duration physical activity along with water intake restriction. A 77 GHz frequency-modulated continuous wave (FMCW) radar is used to capture high-resolution range profiles from the subject's chest, seated in front of the radar, before and after sports activity. Since the electromagnetic penetration into biological tissue is inherently shallow at millimeter-wave frequencies, the sensing is confined primarily to the stratum corneum and upper epidermis. We demonstrate that exercise-induced thermoregulation, perspiration residue deposition, and transient shifts in superficial tissue hydration produce measurable alterations in quasi-static radar reflectivity. We utilize single-chirp snapshots of duration 102.4 micro sec in order to suppress confounders such as slow-varying cardiopulmonary motion, followed by lightweight feature extraction via the Wavelet Scattering Transform (WST) and classification using light-weight machine learning models suitable for edge nodes. Leave-one-subject-out cross-validation on a dataset of 15 sportsmen measured pre- and post-exercise yields a mean classification accuracy of 91.55% with a 95% confidence interval of [84.27, 98.83]%. Interpretability analysis reveals that low-order scattering coefficients (capturing overall signal energy and amplitude stability) dominate discriminative power, consistent with quasi-static changes in skin surface permittivity. The results establish a proof-of-concept for radar-based, edge deployable classification of post-exercise skin states, opening new avenues for non-contact physiological monitoring in sports science.
发表机构
- King Abdullah University of Science and Technology（阿卜杜拉国王科技大学）
机构由 AI 辅助整理，请以论文原文为准。
