Speech extraction based on blind source separation and multi-talker status tracking in automobile environment

The TRINICON (Triple-N ICA for convolutive mixtures) algorithm is utilized to extract the desired speech based on microphones distributed in an automotive cabin.Both the initialization of the TRINICON filters and the multi-talker status determination are based on the spatial relationship between the speech sources and the microphones.Simulations using the captured signal in real automobile environment demonstrate that the proposed method can effectively improve the convergence speed of the TRINICON algorithm.Furthermore,a reliable mapping between the input and output of the blind source separation system is achieved,based on which the desired speech signal can be extracted efficiently.This research provides an effective solution for extracting desired speech in complex automobile environment.