Dereverberation and localization using adaptive reverberation cancellation in the spherical harmonic domain

A dereverberation and localization approach is proposed to improve speech quality and performance of localization of sound sources in reverberant environments,which is based on adaptive reverberation cancellation in the spherical harmonic domain.First,we propose an adaptive reverberation cancellation method based on multi-channel linear prediction in the spherical harmonic domain.Then,for different spherical harmonic orders,dereverberation methods are designed separately using a rigid spherical array and the dereverberated signal is further applied for localization.By using a 32-channel spherical microphone array,simulation results show that compared with the traditional dereverberation algorithm in the spherical harmonic domian,the proposed approach can reduce up to 2/3 computational complexity,as well as significantly improving the PESQ score and SRMR.Experiments in a reverberant room confirm the validity of the proposed method in terms of dereverberation and localization.