應用於即時人體呼吸特徵萃取之低複雜度超寬頻雷達訊號處理系統

A Low-Complexity UWB-Radar Signal Processing System for Real-Time Human Respiratory Feature Extraction

邱于芳(Yu-Fang Chiu)

指導教授 ： 黃元豪

國立清華大學/電機資訊學院/電機工程學系/碩士(2013年)

摘要

此論文提出了可在超寬頻雷達系統擷取人體呼吸訊號中，分析且萃取呼吸訊號特徵的演算法，並且real time實作在訊號處理系統的平台上。在傳統的雷達偵測呼吸訊號的演算法中，大多以偵測出呼吸頻率為主要目標，並將偵測到的頻率用在醫療診斷中。然而在雷達偵測出的人體呼吸訊號中，除了呼吸頻率外，事實上含有更多可以提供醫護人員做為診斷的資訊。因此我們提出了Four Segment Linear Wave的訊號模型來模擬原始呼吸的波形，使用了early stop correlation的方式來取得model中的參數，而這些參數的意義表示了呼吸波形中除了頻率的其餘特徵，例如呼吸的強度、吸氣的狀態、呼氣的狀態、以及呼氣吸氣間轉換的停止狀態。另外，將呼吸model化並取得呼吸特徵參數也可視為壓縮後的呼吸訊號，可以應用在遠端照護監視系統中當作傳輸的標的，如此一來，在不失原本呼吸趨勢的情況下，要表示一周期的人體呼吸可以只用些許的參數來取代至少上千點的取樣資料，明顯的很大的減少需要儲存資料量，也就是使用次方法後可以減少傳輸的頻寬，以及節省傳送的電力消耗。最後，我們此演算法以硬體實現，並透過FPGA邏輯晶片來結合超寬頻雷達系統完成real time的系統驗證。此裝置可適用於呼吸頻率介於0.1Hz到1Hz間的人體呼吸訊號，對於每個呼吸週期分析及萃取。

關鍵字

超寬頻雷達 ； 呼吸訊號 ； 呼吸特徵 ； 即時訊號處理

並列摘要

This paper presents a real time ultra-wideband (UWB) impulse-radio radar signal processing platform with reduced complexity. This platform is integrated with a radar
front-end chip for human respiratory feature extraction and signal compression. Conventional radar detection algorithms only extract respiration rate for medical diagnosis. However, there is more useful information in the radar-detected respiratory signals for medical diagnosis. Thus, this study proposed a four segment linear wave model and an
iterative correlation search algorithm to extract more respiratory features, such as inspiration and expiration speed, respiration intensity, and respiration holding ratio between inspiration and expiration. Moreover, since the iterative correlation search algorithms involves high computation cost, this study applies an early termination scheme and down sampling to reduce the complexity with negligible performance degradation. The extracted features are also useful in a remote medical monitoring system because they can be regarded as compressed respiratory signals. One-period human respiratory cycle can be expressed by extracted features instead of lots of samples. Transmission bandwidth or storage capacity can be greatly saved by transmitting or storing the extracted
features. The proposed algorithm and architecture was designed and implemented on a real time radar signal processing platform with a FPGA chip. Human respiratory
signals from 0.1 to 1 Hz rate are detected and analyzed along with other information in each period.

並列關鍵字

UWB radar ； respiratory signal ； respiratory feature ； real time signal processing

參考文獻

[1] S. Chang, M. Wolf, and J. Burdick, “Human detection and tracking via ultra-wideband (uwb) radar,” in Proc. IEEE Robot. Autom. Conf., may 2010, pp. 452–457.

[2] T.-S. Chu, J. Roderick, S. Chang, T. Mercer, C. Du, and H. Hashemi, “A short-range uwb impulse-radio cmos sensor for human feature detection,” in ISSCC Dig. Tech. Papers, Feb. 2011, pp. 294–296.

[3] C. Caro and J. Bloice, “Contactless apnoea detector based on radar,” in Lancet, vol. 2(7731), Oct. 1971, pp. 959–961.

[4] B. B. C. Franks and D. Johnston, “Contactless respiration monitoring of infants,”in Med. Biol. Eng, vol. 14(3), Oct. 1976, pp. 306–318.

[5] B. B. C. Franks, J. Watson and E. Foster, “Respiratory pattems and risk of sudden unexpected death in infancy,” in Arch. Dis. Child, vol. 55(8), Aug. 1980, pp. 595–604.