Speech Enhancement Algorithm Based On Microphone Array

In our daily life, speech communication is often encountered with noise in cars, shopping malls, streets etc. In some worse environment such as in the war, voice is often submerged in noise generated by explosion. Otherwise, in other fields, such as speech recognition field, the universal application of voice control, voice translated to text and other new technologies which can work properly only in the minimal noise environment are extremely sensitive to the noise. Therefore, the speech enhancement processing technology is of great significance.In early studies, the speech enhancement is mainly based on a single channel of microphone. The single microphone speech enhance algorithm has good noise reduction effect at high SNR environment. However, in low SNR condition, the single microphone algorithm may lead to a substantial reduction of the output speech signal intelligibility. Later, speech enhancement algorithms based on multi-microphone array are proposed. With respect to a single microphone noise reduction algorithm, the multi-microphone array algorithms also make use of spatial information. The multi-microphone array algorithms can make a beam in the direction of the desired speech signal with spatial selectivity to suppress directional interference. Therefore, the multi-microphone array algorithms can be applied to speech enhancement in all kinds of noisy environments.In this paper, the speech communication environment is divided into simple environments and complex environments based on the complexity of the received noisy signal. In the simple acoustic environment, we first elaborated the microphone array speech enhancement based on the generalized sidelobe canceller structure. Then an independent component analysis(ICA) algorithm is used to improve the conventional GSC algorithm. The improved algorithm has a significant speech enhancement effect in a simple environment, but in complex environments(e.g. indoor enclosed environment) the speech enhancement effect is severely reduced. For complex acoustic environment, we focused on the general transfer function generalized sidelobe canceller(TF-GSC) algorithm. This algorithm which is put forward by Shannot combines the generalized side lobe cancellation algorithm with the transfer function(TF) between the signal tothe array. The TF-GSC algorithm has shown impressive noise reduction and low speech distortion abilities in a directional noise field in the complex environment, but less significant noise reduction is obtainable in a diffused noise field. In this paper we combined the TF-GSC microphone array speech enhancement algorithm with a Wavelet wiener filtering method. The Wavelet Wiener filtering method includes with Wavelet Thresholding Multitaper Spectrum algorithm based on the traditional Wiener filtering method. So it has a better noise reduction ability. The discussion is supported by an experimental study using noisy signals recorded in an actual room acoustics environment. Simulation results show that the proposed microphone array speech enhancement algorithm can effectively remove the spectrum of noise in the noisy speech signal, and the output signal has a small distortion.