Research Of Active Noise Cancellation Earphone Algorithms

With the popularity of automobile, research on how to attenuate car interior noise becomes one of the focuses in the realm of signal processing. Active noise cancellation earphone is fit for this application. The prototype of this kind of earphone is Adaptive Active Noise Control (AANC) system, which bases on the theory of Adaptive Filtering. Adaptive filter cancels primary noise by adjusting filter coefficient automatically to emit secondary noise with same amplitude but inverted phase. AANC system is classified into two types by its objective. One which has the only target at reducing noise has been widely used, while the other one, which aims at speech communication as well as noise cancellation, has unsatisfactory performance since one of the requirements could hardly be fulfilled in practice.Having an introduction of LMS adaptive filtering theory and an overview of two types of AANC system, this dissertation proposes two solutions for the purpose of speech communication in car noisy environment. One is to cascade the second type AANC with signal-noise separation system. Three kinds of prevailing algorithms of signal-noise separation are investigated respectively. Noise spectral subtraction has advantages of simple theory and calculation, excellent performance and capability of estimating current noise so that it is selected to apply in the proposed system. Since the spectrum of car interior noise are not constant, dynamic updating of the noise spectral model is required in practical application. For that reason, modified VAD method is utilized to partition the input signal into"speech active segment"and"speech inactive segment". Latest noise spectral model is calculated in speech inactive segment and then employed in spectral subtraction and adaptive filtering in speech active segment. In the end, the pure speech signal is obtained. Another proposed AANC system is combined with spectrum shifting. It makes use of that the car interior noise is at low frequency and separates the speech and noise in frequency domain. This system also achieves the goal of speech de-noise and communication.Description and analysis to these two proposed AANC systems are conducted in details. The performances of them are verified in experiments and the results are the same as expectation.