麥克風陣列實踐空間高解析度毫米動態偵測－以生理訊號為例;Super-resolution Sonar Imaging Using Sparse Microphone Array - A Feasibility Study of Vital Sign Sensing from simulation to reality

NCU Institutional Repository > 生醫理工學院 > 生物醫學工程研究所 > 博碩士論文 > Item 987654321/94791

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/94791

题名:	麥克風陣列實踐空間高解析度毫米動態偵測－以生理訊號為例;Super-resolution Sonar Imaging Using Sparse Microphone Array - A Feasibility Study of Vital Sign Sensing from simulation to reality
作者:	張曦文;Chang, Hsi-Wen
贡献者:	生物醫學工程研究所
关键词:	波束成型;麥克風陣列;聲學成像;點擴散函數;FISTA演算法;frequency domain delay and sum beamforming;microphone array;acoustic imaging;point spread function;FISTA algorithm
日期:	2024-07-27
上传时间:	2024-10-09 15:30:34 (UTC+8)
出版者:	國立中央大學
摘要:	綜觀動態偵測技術的發展，其應用場域不僅適用於姿勢捕捉和跌倒感測等人體動作和位置，還可用於監測生理訊號。在常見的生理訊號中，呼吸為偵測人體健康的一項重要指標，對於疾病的預防與診斷有很大的幫助。現今呼吸監測的方法多數需要使用穿戴式裝置，長時間連續監測下會因為活動空間受限導致行動不便外，也會產生皮膚刺激等問題。非接觸式呼吸監測的發展除了有益於進行長時間的監測，對於遠距醫療與居家檢測的助益也不可忽視。本研究目的在於利用麥克風陣列結合傳統波束成型與反卷積演算法的高解析度成像系統，應用於非接觸式呼吸波形偵測與動作偵測。麥克風陣列應用於聲源定位技術是利用多個麥克風同步收集聲源訊號，透過分析聲波到達的時間和相位差異，將空間中的聲源可視化，實現聲源位置的精確定位。在常規聲源成像的演算法中，最常使用的為延遲求和波束成型法。傳統延遲求和波束成型可以視為一種空間濾波器，凸顯目標方向的聲源並濾除其他方向的聲音干擾。然而，陣列中的麥克風間距與聲源頻率會影響旁瓣的大小與波束帶寬，定位的準確度與成像解析度也會因此受限。本研究開發了一套非接觸式呼吸波形擷取與成像系統，使用二維十六通道麥克風陣列，並以中心頻率為16kHz之頻率調變連續波(frequency modulated continuous wave, FMCW)做為發射訊號，基於波束成型的聲學成像方法，將目標物反射之FMCW訊號聚焦在與陣列平行的XY平面。波束成型之成像可以視為未知聲源與點擴散函數(point spread function)卷積的結果，本研究利用麥克風陣列與聚焦平面之幾何距離建立點擴散函式，並使用fast iterative shrinkage-thresholding algorithms(FISTA)演算法求解反卷積的問題，取得高解析度成像並提高動態擷取之準確度。 ;The development of motion detection technology extends its applications beyond capturing movements such as posture and fall detection to include monitoring physiological signals. Among these signals, respiration is a crucial health indicator that significantly aids in disease prevention and diagnosis. Current methods for respiratory monitoring often rely on wearable devices, which can be inconvenient for prolonged use due to restricted mobility and skin allergy. The advancement of non-contact respiratory monitoring not only supports long-term surveillance but also contributes to telemedicine and home diagnostics. This study aims to utilize a microphone array combined with traditional beamforming and deconvolution algorithms in a high-resolution imaging system for non-contact respiratory waveform and motion detection. Microphone arrays used in sound source localization technology synchronize multiple microphones to collect sound source signals. By analyzing time and phase differences of sound waves, the spatial location of sound sources can be visualized, achieving precise localization. The most commonly used algorithm in conventional sound source imaging is delay and sum beamforming. Traditional delay-and-sum beamforming acts as a spatial filter, highlighting the direction of the target sound source while filtering out noise from other directions. However, spacing in the microphone array and the frequency of the sound source affect sidelobe levels and beamwidth, thereby limiting localization accuracy and imaging resolution. This study developed a non-contact respiratory waveform capture and imaging system using a 2D 16-channel microphone array. We utilize Frequency Modulated Continuous Wave (FMCW) signals with a center frequency of 16kHz as the transmission signal. Based on beamforming acoustic imaging, the reflected FMCW signals from the target are focused on the XY plane parallel to the array. Furthermore, imaging via beamforming can be considered as the convolution of an unknown source and the point spread function (PSF). This research establishes the PSF using the geometric distance between the microphone array and the focusing plane and employs the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) to solve the deconvolution problem, achieving high-resolution imaging and enhancing motion capture accuracy.
显示于类别:	[生物醫學工程研究所 ] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	27	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....