Noise Suppression with High Speech Quality Based on Kalman Filter

We propose a noise suppression algorithm based on Kaiman filter. The algorithm achieves a noise suppression with high speech quality under the condition of the AWGN (additive white Gaussian noise), from the canonical state space models with a state equation composed of the clean speech signal and an observation equation composed of the clean speech signal and AWGN. The special feature of the proposed algorithm is realization of high speech quality noise suppression utilizing only the Kalman filter, while the conventional algorithm utilizes the linear prediction algorithm and the Kalman filter. The simulation results show that the proposed method improved the noise suppression capability by about 5 dB than that of the conventional method