Improve equalization-cancellation-based sound localization in noisy reverberant environments using direct-to-reverberant energy ratio

摘要

We previously proposed an algorithm for binaural sound source localization based on the equalization-cancellation (EC) binaural hearing model. Though this sound source localization approach exhibits relatively good results in noisy conditions, its performance in the presence of reverberation dramatically degrades. To deal with this problem, in this paper, the EC procedures on which the sound localization approach was designed are firstly analyzed. Subsequently, we propose to further improve the previous sound source localization algorithm in reverberant conditions by adapting the parameters of the EC model to the present conditions based on the direct-to-reverberant energy ratio (DRR). Experimental results demonstrate that the improved sound source localization algorithm outperforms our previously-proposed and other traditional sound source localization algorithms under noisy reverberant environments.