Least relative entropy for voiced/unvoiced speech classification

The aim of this work is to develop ajlexible and eficient approach to the classifcation of the ratio of voiced to unvoiced excitation sources in continuous speech. To achieve this aim we adopt a probabilistic neural network approach. This is accomplished by designing a multi layer perceptron classifer trained by steepest descent minimization of the Least Relative Entropy W) cost function. By using the LRE cost function we can directly ou@ut the ratio, as aprobabiliv, of excitation source, voiced to unvoiced, for a given speech segment. These output probabilities can then be used directly in other applications, such as low bit rate coders.