Artificial intelligence

Abstract: Wavelet-analysis based method for speech boundaries detection in a noised signal was offered. As one of stages this method includes input signal’s frames classification, which is based on wavelet spectrum energy characteristics. It allows to take into account acoustic characteristics of speech sounds’ wide classification. Such an approach gives an opportunity to allocate a speech in a signal with high-amplitude noises, to execute a speech signal segmentation and to raise efficiency of further recognition.

Keywords:

References: