Active Noise Control with Speech Enhancement Using Inverse Notch Filter

Abstract

This paper proposes an active noise control (ANC) system based on speech enhancement using inverse notch filter (INF). Conventional ANC systems remove an observed signal, even if it includes a desired speech signal. To extract the desired speech signal, we introduce a speech reduction method as pre-processing of the ANC system. In this case, the ANC produces only the noise replica, since the input signal to the ANC does not include the speech signal. We inversely use a speech enhancement technique to achieve speech reduction. One of the useful speech enhancement techniques is a spectral gain method. Inherently, the spectral gain method based on the FFT provides a delay depending on the analysis frame length. Since the ANC system can work well under a low processing delay, we replace the FFT with the INF which is a narrow band-pass filter for reducing processing delay. Simulation results show that the proposed ANC system with INF is effective to extract the speech signal included in the observed signal.