應用延遲濾波器之麥克風陣列訊號處理

Abstract

本論文主要針對密閉空間之迴響環境，進行延遲加成波束合成器於訊號源方位判別與空間濾波之可行性探討。本論文亦提出一個符合實驗室規模、具彈性、16通道之即時麥克風陣列訊號處理實驗平台，並已實作完成。在此實驗平台上，吾人除了錄製若干麥克風陣列語音樣本供相關研究用外，也實作完成以延遲加成波束合成器為基礎之即時訊號源方位判斷及空間濾波器。 此外，本論文亦進行響應補償波束合成器之可行性探討，並與延遲加成波束合成器比較兩者之成效。

In this thesis, an investigation is done for the performance of DOA estimation and spatial filtering based on a delay-and-sum beamforming using fractional-delay FIR (FD-FIR) filters in a reverberant environment. Besides, a flexible real-time array-processing platform of 16 sensors is proposed and implemented. Using this platform, some speech patterns are recorded for related off-line experiments. Real-time DOA estimator and spatial filter are also implemented. Further more, the performance of model-inversion beamforming is investigated and confronted with the above delay-and-sum beamforming using FD-FIR filters.