Energy-based voice activity detection algorithm using Gaussian and Cauchy kernels
Aminadabe dos Santos Pires Soares, Wemerson Delcio Parreira, Everton Granemann Souza, Sérgio José Melo de Almeida, Cláudio Diniz, Chiara das Dores do Nascimento, Matheus F. Stigger · 2018
In this work we present a simple and robust Energy-Based Voice Activity Detection Algorithm using Kernel (KVAD). Taking advantage of kernel metrics, Gaussian and Cauchy kernels are used to classify acoustic signatures as speech and non-speech. As an evidence of the potentiality of KVAD algorithm, comparisons with existing energy-based algorithms are presented, showing better performance in adverse environments as low signal-to-noise ratio (SNR) and non-stationary noise.