聲源方向辨識系統研製

Abstract

本論文設計及研製一哨子聲源方向辨識系統，並探討於機器人操控之應用。此系統之偵測角度範圍為360°，辨識誤差為±5°。語音辨識提供智慧型人機介面一項重要的技術，但在吵雜環境中難以保持良好的辨識結果。本論文發展之哨子聲源方向辨識系統，可改善上述環境對系統造成之影響，正確工作於吵雜環境。一般聲源方向辨識之解碼方式複雜，需借助電腦運算方能計算出聲源方向，限制了系統的應用。本論文提出一簡易解碼方式，可由電路得知聲源方向，提高系統可攜性。因應未來發展，本論文提出一混合式晶片(Hybrid chips)聲源方向辨識系統之架構，結合陣列矽晶麥克風與聲源辨識晶片，逹成一智慧型感測器。並將所設計之哨子聲源方向辨識系統晶片化，大幅縮小系統體積及功率需求。

In this thesis, we develop a whistle sound direction detection system. It can detect sound direction with an accuracy of ±5°. Most speech recognition systems cannot keep their performance in a noisy environment. However, the developed whistle sound detection system can surmount this problem. More over, we integrate the system to mobile robots. Experimental results show that the robots can be controlled using a whistle in a noisy environment. A new decode method is proposed in this thesis to resolve the sound direction. This method uses circuit to obtain direction information, an improvement to the conventional computer dependent methods. We propose a hybrid chip approach to a integrated system. In this system, silicon microphone array is combined with a sound direction detection chip to facilitate future application.