An improved iterative wiener filtering algorithm for speech enhancement

Speech enhancement is a critical technique for various applications like mobile communication and automatic speech recognition (ASR). This paper studies an improved iterative Wiener filtering (IWF) algorithm, which can help enhance the recognition rate of the ASR system. By using the voice activity detection (VAD) technology in the traditional IWF algorithm, the noise power spectrum estimation in the silent periods can be improved. Besides, the mini-tracking algorithm is also employed to estimate the signal-to-noise ratio (SNR) of the present speech signal, which is further used to control the number of iterations for the IWF algorithm. Therefore, the speech-like properties which are important in speech reconstruction and recognition are better retained. Computer simulations are conducted to verify the proposed algorithm has improved performance in ASR system over the conventional approach.