Comparative Study on Voice Activity Detection Algorithm

This paper discussed several algorithms of the voice activity detection (VAD) in detail based on the feature coefficients extraction and analysis between voice signal and noise, including short-time energy, short-time average magnitude function (AM), short-time average zero-crossing rate (ZCR), short-time Auto Correlation, short-time average magnitude difference function (AMDF) in time domain, and Fourier analytics in frequency domain. Moreover the paper illustrated experiment simulation of these algorithms and comparative analysis of simulation results. The results show that each algorithm has its own advantages, and voice detection has disadvantage in certain areas by using an algorithm individually. It can maximize the advantages of each algorithm and improve signal detection accuracy by combining several algorithms learn from each other to build a multi-level VAD system.