利用空間域特徵空間一致性及共鳴曲線相似度之喚醒關鍵字偵測方法

Abstract

本論文提出了一套使用麥克風陣列偵測喚醒關鍵字的方法，本方法運用聲源於麥克風陣列在不同頻率下的特徵空間一致性(Spatial Eigenspace Consistency)，以及喚醒關鍵字語音共鳴曲線相似度(Resonant Curve Similarity)作為判別關鍵字的特徵，並藉由貝氏風險評估與串聯式偵測器的結合做為偵測的機制。此方法在極低訊噪比下仍保有相當強健的辨識率，因而可以適用在遠距關鍵字語音偵測或者在吵雜的環境下作為關鍵字語音喚醒機制。除了偵測關鍵字外，本方法還能同時估測出關鍵字的聲源方向，並保有串接其他偵測器的能力。在有額外的語音特徵或空間特徵可以加入時，能夠簡易的設計新的偵測器，串接到原本的架構上以持續增進辨識率。

This thesis proposes a method to detect keywords using microphone array. The consistency of the spatial eigenspaces formed by the speech source at different frequencies and the resonant curve similarity of the keyword are used as the features for detection. These features are processed and detected separately and the result is determined by cascading individual outcome using Bayes risk estimate. This proposed method can keep a high recognition rate under very low SNR conditions. Therefore, it is suitable for the applications such as distant wake-up and keyword detection in noisy environments. In addition, this method can estimate the direction of arrivals of the sound source, and the proposed architecture is easy to expand by adding other feature detection methods in the cascaded manner to further improve the recognition rate.