Speech noise reduction using sequential spectrum detection based on modified DFT pair

A single-channel speech noise reduction using sequential spectrum detection based on a modified DFT pair is proposed in this paper. In the proposed method, a noise added speech is decomposed into frequency signals using the modified DFT and then signal and noise elements are sequentially detected from the decomposed frequency signal. Noise subtraction is operated by subtracting the noise element from the signal one. However, it needs a coefficient for compensating the difference between the detected noise level and the proper one. Therefore, an automatic setting of the coefficient is adopted in the noise subtraction. The effectiveness of the proposed noise reduction method is confirmed by subjective evaluations using processed signals.