以轉移函數比值之方法實現雙聲道麥克風聲源分離系統

Abstract

本論文針對環境中的語音訊號干擾源：如喇叭回音、背景音樂、或非使用者的語音等，使用雙聲道麥克風訊號處理，抑制干擾源的影響，保留使用者的語音資料而達到聲源分離的目的。本系統在硬體上使用Microchip推出的PIC18F4550來收集雙聲道類比麥克風資料並將其轉換為數位訊號，再透過USB2.0介面將聲音資料傳遞至電腦端，最後利用轉移函數比值演算法(Transfer Function Ratio)做出空間濾波器，達到聲源分離的效果。可以應用在吵雜環境中的語音純化等方面。

The objective of this thesis is to suppress the interference signals such as loudspeaker echo, background music, or competing speech. We design an array signal processing technique on a dual-channel microphone for sound sources separation, which minimizes the influence of interference sources and retain the target speech. This system collects the audio data from two analog microphones via Microchip PIC18F4550, transforming the audio data into digital data, and transmits to PC through USB 2.0 interface. We implement the spatial filter by transfer function ratio approach for sound sources separation. Experiments for different types of interferences are conducted to show the effectiveness of the proposed system.