Noise Reduction in Speech Signals Using Discrete-time Kalman Filters Combined with Wavelet Transforms

Noise reduction in speech signals is a growing area that encountered several applications like communication channel transmission, automatic speech recognition, telephony and hearing aids, among others. This paper introduces a technique for noise reduction in speech signals that combines both Discrete-time Kalman filtering and Wavelet transforms. While filtering provides noise reduction, Wavelets transforms allow minimizing spectral distortion. In order to assess the efficiency of this combination, we compared both the segmental signal-to-noise ratio and the Itakura-Saito distance at the input to their respective values at the output of the proposed system. Also, we compared the noise reduction performance of the proposed system to that of Kalman filtering and the combination of Wavelet transforms with Kalman filtering has shown satisfactory results.