陣列麥克風系統實現與MEMS電容式麥克風設計

Abstract

電容式麥克風具有較高的靈敏度與較平坦與寬廣的頻率響應，所以廣泛的應用於各種場合。結合陣列訊號處理技術可提高陣列麥克風的訊噪比(Signal-to-Noise Ratio)與指向性(Directivity)。在此，使用超指向性法(Superdirective method) 來達成聲束形成(Beamforming)。超指向性法顧名思義，可使聲束指向前方且比一般的聲束為窄，其作用於低頻時更加顯著。能有效提高低頻指向與訊噪比。實現上，利用polyphase進行升頻(up-sampling)與降頻(down-sampling) 減低運算量，再加入單一通道的Wiener post-filter來有效抑制噪音，以及一adaptive filter (NLMS) 將回聲(echo)消除。最後，使用SRP-PHAT將吵雜環境中的聲源定位 (Direction of Arrival)，提升接收的清晰度。 接著試著將麥克風縮小，設計製程步驟，利用微機電技術將體積微小化，使其有最佳的頻寬與靈敏度。再將微麥克風結合於陣列訊號處理上，達到最佳的效果。期望此一系統可應用於手機、助聽器、或車內免持聽筒上。

Condenser microphones have higher sensitivity with flat and wide frequency response, so that they are commonly used in different applications. By integrating the array signal processing, the SNR (Signal-to-Noise Rate) and the directivity can be optimized. In this thesis, we use superdirective method to do beamforming. Superdirective method is to narrow down the beamwidth and make it pointing to the front directly. The effect is more obvious while working in the low frequency, and can dramatically improve the high/low frequency directivity and SNR. Additionally, process the array output by a single-channel noise suppression filter (Wiener post-filter), and then eliminate echo with an adaptive filter (NLMS). Finally, use SRP-PHAT to do DOA in reverberant rooms. Therefore, it can be utilized for receiving signal clearly to enhance SNR and directivity. Also, we minimize the microphone volume by Micro-Electro-Mechanical System (MEMS) technology, in order to get the best bandwidth and sensitivity. Then combine the MEMS condenser microphone with the array signal processing, trying to achieve the best quality. In the future work, this system could be used in the cellular phone, hearing-aid, and the car mobile applications.