A METHOD AND A SYSTEM FOR DETECTING VOICE ACTIVITY BASED ON A COMPLEX GAMMA STATISTICAL MODEL

摘要

A method and an apparatus for VAD based on a complex gamma statistical model are provided to use the complex gamma statistical model instead of an existing Gaussian statistical model, thereby improving VAD performance according to types of noise and SNR(Signal to Noise Ratio) conditions. A method for VAD(Voice Activity Detection) based on a complex gamma statistical model comprises the following steps of: converting an inputted voice signal into a voice signal of a frequency domain by performing FFT(Fast Fourier Transform) of the inputted voice signal; estimating power of a noise signal from the voice signal converted into the frequency domain; deciding a likelihood ratio on the frequency domain on the assumption that a DFT(Discrete Fourier Transform) coefficient of clean speech and noise follows the complex gamma statistical model with respect to existence and nonexistence of the voice based on the estimated power of the noise signal; and calculating a decision rule for the VAD from the decided likelihood ratio.