Detection of human speech in structured noise

摘要: This paper describes research to develop an efficient system that provides a binary decision as the presence of speech in short (one three second) time sample acoustic signal. A method which is and reliably detects human structured noise (such wind, music, traffic sounds, etc.) described. Two separate algorithms were developed. The first algorithm by testing for concave and/or convex formant shapes. second statistical pattern classifier utilizing radial basis function (RBF) networks with mel-cepstra feature vectors. Classification errors are not consistent across these two different methods. As consequence, we plan reduce our error rate fusion >