MODEL-BASED SPARSE COMPONENT ANALYSIS FOR REVERBERANT SPEECH LOCALIZATION

This paper studies the problem of multiple speaker localization via speech separation based on model-based sparse recovery. We compare and contrast computational sparse optimization methods incorporating harmonicity and block structures as well as autore-gressive dependencies underlying spectrographic representation of speech signals. The results demonstrate the effectiveness of block sparse Bayesian learning framework incorporating autoregressive correlations to achieve a highly accurate localization performance. Furthermore, significant improvement is achieved using ad-hoc mi-crophones for data acquisition set-up compared to the compact mi-crophone array. Index Terms — Structured sparsity, Reverberant speech local-ization, Autoregressive modeling, Ad-hoc microphone arra