語音強化與立體聲迴聲消除於智慧型電視之應用

Abstract

本論文提出一套利用麥克風陣列結合適應性空間濾波器及聲學迴聲消除的方法進行語音強化。智慧型電視的興起，使遠距語音訊號處理應用再度受到重視，例如語音命令控制和視訊會議都屬於遠距語音的應用，利用這些功能的同時第一個面對的問題即是如何抑制電視播放的聲音、迴聲和環境噪音對目標聲源之影響，這將是語音介面是否能夠成功運用於智慧型電視的關鍵。 本論文設計一套能消除立體聲迴聲並且抑制干擾源和剩餘迴聲之適應性波束形成器。藉由實際智慧型電視的環境配置下進行實驗模擬，麥克風收到訊號分別經過聲學迴聲消除處理後在通過最小方差無失真響應的波束形成器，配合適應性噪音消除達到語音強化的效果；其中空間前處理和噪音估測所需的角度資訊，事先運用轉移函式比值法求得目標聲源在空間中相對轉移函式。

An approach proposed in this thesis combines adaptive spatial filtering of microphone array and acoustic echo cancellation for speech enhancement. The growing trend of smart TV users has reintroduced the value of using long distance speech signal process application again. For example, voice control and video conference both are long distance speech applications. When you use those applications, the one of the problems you may encounter is how to minimize the TV sound, echo, and environment noise all of which affects the target source. This issue is the key to whether or not smart TV’s speech interface will be successful. This thesis presents a stereophonic acoustic echo cancellation (SAEC) after using an adaptive beamformer for noise reduction and residue echo cancellation. In the real smart TV environment setting, this thesis simulated and experiment with, microphones which received voice signals that passed through a stereophonic acoustic echo cancellation process and then a minimum variance distortionless response (MVDR) beamformer combine adaptive noise cancellation (ANC) for speech enhancement. Spatially pre-process and noise estimate both need to know the directional information between target source and microphones in any given space. To find out this data, one must calculate relative transfer function (RTF) using the transfer function ratio (TFR) method.