頭戴型跌倒偵測系統

一、研究目標

跌倒會給老年人帶來心理創傷和身體傷害,且老年人若在跌倒發生後無法被及時救援跌倒,可能會造成嚴重的傷勢,本團隊利用慣性感測器開發多部位穿戴式動作偵測系統,此系統能即時擷取人體頭部與腰部姿態訊號,並藉由人工智慧演算法辨識使用者目前姿態的概況,並在跌倒事件發生後給予及時的辨識與警訊的提醒,此外此系統亦可針對步態資訊進行分析,以建立平衡能力評估與跌倒事件的預測,再搭配醫師在復健動作與生活習慣之建議,將能有效改善高齡者與泛亞健康族群老年生活品質。目前多數跌倒偵測系統,無法正確偵測出在不平坦的地板表面發生的幾種具有高死亡風險之跌倒事件,例如:樓梯或有障礙物的地板,對此本團隊提出了一種將距離感測器嵌入鞋底的跌倒感測裝置,可以識別在樓梯和不平坦地板上的跌倒和近似跌倒的動作。

圖1、多部位穿戴式系統硬體開發

圖2、資料收集流程與演算法開發

二、研究成果

針對本團隊開發之多部位穿戴式系統,頭部裝置以眼鏡外觀配戴於頭上,在日常活動中運動頭部不會發生顯著變化,因此此裝置可以透過跌倒時裝置之均方根加速度、俯仰角和滾動角的變化,配合閾值法進行比較且判斷是否發生跌倒。然而在頭部角度估計上精準度會因感測器本身之誤差而有所影響,為了消除誤差和計算轉動角度,利用加速度計和磁力計的數據及Madgwick 濾波器來校正陀螺儀的漂移,研究成果顯示頭部裝置辨識跌倒的精確度能達到97.75%,不僅降低誤報率,辨識度上也相當精確。此外,為了研究來自頭部的慣性傳感器信號與步行平衡能力之間的相關性,本團隊構建了一個基於深度學習神經網絡的模型,並對因肌肉發育不平衡、神經病變而具有跌倒風險的三種異常步態進行辨識,包含:止痛步態(antalgic gait)、蹣跚步態(waddling gait)、搖擺步態(Trendelenburg gait),通過10米步行實驗捕捉到的異常步態的IMU信號,並引用時間卷積網絡(TCN)模型,實現對三種異常步態的分類。這些步態可能發生在行走時突然出現疼痛或肌肉無力的情況下,使人面臨跌倒危險。分類結果顯示,準確率、召回率、精確率和F1分數的平均值分別為98.33%、98.27%、98.2%和98.23%。在另一項實驗中,受試者在行走過程中從正常步態轉變為異常步態,以驗證異常步態檢測的時間延遲。平均時間延遲為2.52秒,檢測準確率達到100%。整體結果表明,該方法提供了一種通過頭部運動監測步態的新方法。

圖3、眼鏡型穿戴式裝置IMU資料視覺化

圖4、眼鏡型穿戴式裝置步態分析流程

另外基於足部鞋型穿戴式裝置的ToF距離量測訊號,本團隊提出了一種確保估計距離準確性的跌倒偵測方案,此系統將使用ToF 信號中提取的特徵導入 Error-correcting Output Codes-based (ECOC)的分類模型來識別日常生活動作、類跌倒事件與跌倒事件類共17項活動,為了驗證所提出的跌倒檢測方案,從九名參與者的動作中生成了一個包含450個日常生活動作、216個類跌倒事件動作和432個模擬跌倒的實驗數據集,在這三個類別的實驗中,F1 score分別為 100%、98.17% 和 99.07%,此實驗證明本系統可以精確地辨別類跌倒事件與不同地型上的跌倒偵測。 濾波器來校正陀螺儀的漂移,研究成果顯示本裝置辨識跌倒的精確度能達到97.75%,不僅降低誤報率,辨識度上也相當精確。此外,為了研究來自頭部的慣性傳感器信號與步行平衡能力之間的相關性,本研究構建了一個基於深度學習神經網絡的模型,並對具有跌倒風險的步行平衡狀態進行分類。

圖5、鞋型穿戴式裝置ToF訊號視覺化

實驗室相關研究:

[1] Chih-Lung Lin*, Fang-Yi Lin, Cheng-Yi Huang, Yuan-Hao Ho, Wen-Ching Chiu, and Pi-Shan Sung, "Classification of Gaits with a High Risk of Falling Using a Head-Mounted Device with a Temporal Convolutional Network," IEEE Sensors Letters, vol. 8, no. 5, pp. 1-4, May 2024.

[2] Chih-Lung Lin*, Yuan-Hao Ho, Wen-Ching Chiu, Ting-Ching Chu,You-Hong Liu, "Innovative Shoe-Integrated System Based on Time-of-Flight Range Sensors for Fall Detection on Various Terrains," IEEE Sensors Letters, vol. 5, no. 10, pp. 1-4, Oct. 2021, Art no. 6002404.

[3] Chih-Lung Lin*, Wen-Ching Chiu, Ting-Ching Chu, Yuan-Hao Ho, Fu-Hsing Chen, Chih-Cheng Hsu, Ping-Hsiao Hsieh, Chien-Hsu Chen, and Chou-Ching K. Lin, Pi-Shan Sung, Peng-Ting Chen, "Innovative head-mounted system based on inertial sensors and magnetometer for detecting falling movements," Sensors, vol. 20, pp. 1-16, Oct. 2020.

[4] Chih-Lung Lin*, Wen-Ching Chiu, Fu-Hsing Chen, Yuan-Hao Ho, Ting-Ching Chu, and Ping-Hsiao Hsieh, "Fall Monitoring for the Elderly Using Wearable Inertial Measurement Sensors on Eyeglasses," IEEE Sensors Letters, vol. 4, no. 6, pp. 1-4, Jun. 2020.

[5] Cheng-Yi Huang, Yuan-Hao Ho, Yu-Chi Hsiung, Peng-Ting Chen, Pi-Shan Sung, and Chih-Lung Lin*, "Efficient Fall Detection Method Using Time-of-Flight Sensors and Decision Tree Model," 2023 IEEE 12th Global Conference on Consumer Electronics (GCCE), pp. 1118-1119, Nara, Japan, Oct. 2023.

[6] Li-Fan Tseng, Po-Ting Lee, Pi-Shan Sung, Peng-Ting Chen, and Chih-Lung Lin*, "Sit-to-Stand Time Measurement System Based on Eyewear Device for the Elderly," 2023 IEEE 12th Global Conference on Consumer Electronics (GCCE), pp. 208-209, Nara, Japan, Oct. 2023.

[7] Fang-Yi Lin, Po-Ting Lee, Yuan-Hao Ho, Pi-Shan Sung, Peng-Ting Chen, and Chih-Lung Lin*, "Fall Prediction Based on Head-Mounted IMU Sensor System," 2022 IEEE 11th Global Conference on Consumer Electronics (GCCE), pp. 341-343, Osaka, Japan, Oct. 2022.

[8] Meng-Hsuan Wu, You-Hong Liu, Chieh-Hsu Chen, Peng-Ting Chen, and Chih-Lung Lin*, "DTW-based Motion Matching System in Consumer Shopping Interest Analysis," 2021 IEEE 10th Global Conference on Consumer Electronics (GCCE), pp. 606-607, Kyoto, Japan, Oct. 2021

[9] Po-Ting Lee, Wen-Ching Chiu, Yuan-Hao Ho, You-Cheng Tai, Chou-Ching K. Lin, and Chih-Lung Lin*, "Development of Wearable Device and Clustering Based Method for Detecting Falls in the Elderly," 2021 IEEE 10th Global Conference on Consumer Electronics (GCCE), pp. 231-232, Kyoto, Japan, Oct. 2021.

[10] Wen-Ching Chiu, Ping-Hsiao Hsieh, Feng-Ching Cheng, Chieh-Hsu Chen, Chou-Ching K. Lin, and Chih-Lung Lin*, "Implementation of Rider Assistance System with Head-up Display and Safety Notification," 2019 IEEE 8th Global Conference on Consumer Electronics (GCCE), pp. 432-433, Osaka, Japan, Oct. 2019.