基於頻譜形狀敏感度的感知聲學主動式噪音控制系統

Abstract

傳統的主動式噪音控制系統通常都只有從能量的觀點來看殘留噪音的大小，如NAL、MSE這兩種直接由殘留噪音的振福做量化的評比，沒有考慮到人耳的聽覺感知。人的聽覺對不同頻率會有不同的敏感度，除此之外，大腦對聲音的時-頻域調變也有不同的敏感度，我們可以利用這些特性來設計較符合人類聽覺感知的主動式噪音控制系統。

在本論文中，我們除了利用人耳的聽覺閾值設計的預強調濾波器外，還利用大腦對頻率調變的感知閾值設計預強調濾波器。將這兩個預強調濾波器用於主動式噪音控制系統的演算法中，對噪音的不同頻率和不同調變頻率採取不同的權重。我們除了在傳統的主動式噪音控制系統上，另外還在混合、頻域及子頻帶三種主動式噪音控制系統上進行實驗。我們使用白雜訊及多個單頻音來進行模擬，最後計算殘留噪音的響度及進行心理聲學實驗來證明我們利用設計的預強調濾波器來增加敏感頻率和敏感調變頻率的消除，確實會使殘留噪音較為小聲。

Conventional active noise control systems typically evaluated performance using the power of residual noise. They are seldom evaluated from perceptive viewpoints by the residual noise. Human hearing has different sensitivity to different frequency. In addition , the characteristic of that the brain analyzes sounds in terms of spectro-temporal modulation, can also be used to enhance the residual noise audibility .

In this thesis, in addition to using hearing threshold to design pre-emphasis filter, we also use the perception threshold of spectral modulation to design pre-emphasis filter. These two pre-emphasis filters used for noise weighting are incorporated into active noise control systems. Apart from the conventional active noise control system, we also based on hybrid、frequency domain and subband adaptive filtering structures. We use white noise and multi-tone signal to simulate .Calculating loudness of residual noise and listening tests show that our design can improve the performance in noise cancellation.