Temporal Representation of Vowels in Khasi Dialect

Abstract

Some speech parameters such as short-time energy, zero crossing rate, and minimum amplitude of the signal can be estimated directly by temporal analysis. This analysis is the easiest method of representing a speech signal since temporal features are easy to extract and have a very easy physical interpretation. An efficient automatic speech recognition system can be achieved by preserving the information needed to determine phonetic identity of the speech signal. Representation of speech in appropriate form yields a more efficient coding system as it can improve the quality of speech synthesis and also increases the performance of speech recognizers. In this paper, phonemic vowels found in Khasi dialect such as /a/, /i/, /I/, /e/, / ϵ /, / ⊃ /, /o/, and /u/ are represented purely based on temporal features. Depending on the presence of periodicity information on vowel sounds, fundamental frequency (F 0 ) of each vowel has been calculated using autocorrelation function.