利用超寬頻雷達偵測與評估人體呼吸運動

Abstract

本研究的主旨為利用超寬頻雷達偵測人體呼吸運動，而本研究的主要貢獻為超寬頻雷達的信號處理演算法之改良設計及系統整合實現。文獻中已有相關的研究探討利用超寬頻雷達所接收到一維資訊偵測呼吸及心跳，而他們的方式為利用互相關函數(cross-correlation function)或直接利用頻譜分析的方式。但以上方法的問題為效能將會取決於其取樣頻率且皆僅分析一維的資訊，因此無法得知物體在空間中的確切位置。在此研究中，利用自相關函數(auto-correlation function)，改為估計信號的相位，以解除取樣頻率及解析度之間的直接關係。利用此方法所估計的人體呼吸頻率，誤差僅有8%。除此之外，為了要能夠得知物體的二維位置，利用合成孔徑的技術將接收到的超寬頻信號做二維成像。兩種方法：相位校正法(phase-correction method)及同調性因子可適性權重法(generalized coherence factor based adaptive imaging method)，也被應用來增進影像的品質。結果發現此兩種方法的結合將可以同時降低主瓣寬度至原來主瓣寬度之40%，及降低旁瓣強度達20dB。最後則提出能夠結合合成孔徑成像技術並偵測人體呼吸運動的演算法。其偵測原理為利用廣義非同調性因子(generalized incoherence factor)：在物體有呼吸運動的區域，其廣義非同調因子的值將較高。加以利用濾波器組(filter bank)的技術並定義濾波器組廣義同調性因子(filter bank based generalized coherence factor)以分析信號頻域特性，將可以同時評估呼吸運動的頻率。由廣義非同調性因子演算法所得到的影像強度，分別在人體有呼吸運動及沒有呼吸運動的區域，其比例約接近33 (30dB)；而使用濾波器組廣義同調性因子演算法的結果，在與人體呼吸頻率不同/相同的濾波器中所得到的能量比例為8.33 (18dB)。

The purpose of this research is to use Ultra-wideband (UWB) Radar for human respiratory motion detection and the main contributions of this study are the design of UWB radar signal processing algorithms as well as its system implementation. Prior researches in the literature on the use of UWB radars for human respiratory motion detection are mainly based on cross-correlation functions or spectrum analysis. These methods typically require high sampling frequency and often only capable of 1D measurements. In this study, the auto-correlation function is used instead in order to achieve higher resolution and alleviate the limitation on the sampling rate. The estimation error of the breathing rate is 8%. In addition, the synthetic aperture technique is also used for 2D imaging. Two methods, including the phase-correction method and the generalized coherence factor method, have been applied for image quality improvements. And it was found that the combination of the two methods can reduce the mainlobe width to 40% of its original width and reduce the sidelobe level by 20dB. Finally, an algorithm that combines the synthetic aperture technique and human respiratory motion detection was proposed. The algorithm makes use of the generalized incoherence factor (GICF), and the value of GICF increases at the region where there is human respiratory motion. In addition, the respiratory motion rate is estimated by defining filter bank based GCF (FBGCF). The image intensity at regions with and without breathing human has a ratio of 33 (30dB). And for the FBGCF method, the ratio of the energy at the filter that corresponds to human respiratory motion and the energies at the other filters is 8 (18dB).