Speech segmentation and feature normalization based on area functions

This paper presents a new approach to automatic segmentation and feature normalization of connected speech based on area functions. Algorithms for segmenting speech sounds into vowel-like and nonvowellike segments, and then for identifying vowels and detecting nasal segments, turbulence noise segments, etc. are described, together with an algorithm for feature normalization. Fairly reasonable results were obtained with seven sentences spoken by two male speakers and a female speaker.