Rayleigh providers robust voice activity detection
Wang Jingfang, Huiyan Xu · 2011
In this paper, the low signal to noise ratio, non- stationary noise environment speech endpoint detection, a business based on Toeplitz of Rayleigh Noise endpoint detection methods. The method uses multi-frame signals adjacent to construct a symmetric Toeplitz covariance matrix, the Rayleigh advantage of the Toeplitz operators of double the amount of information on the voice signal endpoint detection threshold. New algorithm has been tested to effectively distinguish between speech and noise, low-noise in different environmental conditions has good robustness. And other methods, the accuracy is higher. For speech feature extraction and recognition provides a new way.