Blind Source Separation for Intelligent Vehicles Based on Microphone Array in Road Environment
Chao Sun, Sifan Wang, Qi Lecky Li · 2021
Compared with optical signal, sound signal is endowed with advantages of cheaper sensor, less blind area and non-visual field perception. The application of sound perception in intelligent vehicles can enhance the reliability of environment perception, but the problem of blind signal separation in traffic environment should be solved first. In this paper, an improved Fast Independent Component Correlation (Fast-ICA) algorithm is applied to the scene of road delay signal mixing to realize blind source separation of sound signal in the road environment. Firstly, Fast-ICA algorithm is extended to the complex domain to process the sound signal in time and frequency domain. Then, the pre-processing and post-processing methods are proposed based on the road environment. The results of the experiments and simulation show that the extended Fast-ICA algorithm has good adaptability to the time-delay characteristics of road environment, and can effectively separate the sound sources of main sound signals, and provide high-precision sound source signal input for acoustic-based positioning method.