An efficient speech enhancement method using filter and spectral subtraction

This work presents an efficient method to enhance speech. The external noise is eliminated and the clean speech is estimated by combining the features of the Kalman filter with the spectral subtraction technique. Compared with conventional spectral subtraction algorithm, our method can reduce the effect on the spectrum estimates, which disturb not only the listener but also speech coding algorithms. The proposed method has been tested via many experiments. The results demonstrate the effectiveness of the method.