A Study on Noise Reduction Method Based on LPEF and System Identification with Step Size Control

We have proposed a noise reduction method based on a noise reconstruction system (NRS). The NRS uses a linear prediction error filter (LPEF) and a noise reconstruction filter (NRF) which estimates background noise by system identification. In case a fixed step size for updating tap coefficients of a NRF is used, it is difficult to reduce background noise while maintaining the high quality of enhanced speech. In order to solve the problem, the control of a step size is proposed. It takes advantage of cross-correlations between input signals and an enhanced speech signal. In a speech section, a small step size is used so that speech cannot be estimated. On the other hand, a large step size is used to track non-stationary noise in a non-speech section