A Study on Step Size Control for Noise Reconstruction System with ALE

Abstract

A noise reduction method is proposed to reduce background noise in noisy speech. We have investigated a noise reconstruction system (NRS) with an adaptive line enhancer (ALE). The NRS is based on a linear prediction error filter (LPEF) and a noise reconstruction filter (NRF). An output signal of a LPEF becomes a white signal. Assuming that a white signal generates background noise by exciting a linear system, a NRF can reconstruct the background noise from white noise by estimating the linear system. However, the estimation accuracy of background noise decreases, when residual speech is included in an input signal of a NRF. Thus the ALE is used to decrease the residual speech. In this paper, in order to improve the noise reduction ability more, a noise reconstruction system using step size control is proposed. The step size control makes use of cross-correlations between input signals and an enhanced speech signal.