Recognition of Arabic phonemes using fuzzy rule base system

Abstract

Speech is the most common and efficient mode of communication among humans. Speech recognition is one of the most complex pattern matching problems. The objective of speech recognition is to design a system that can recognize speech from multiple speakers with variability in speech rate, pitch, volume etc. and from a large vocabulary. The basic components of speech are phonemes and recognition of words of some language depends on the phonemes of individual letters constructing that word. This experiment is about the recognition of 28 basic phonemes of the Arabic language. The recognition system is a fuzzy based system. The input to the system is in the form of frequency intensity of the phonemes. Fuzzy rules are defined on the basis of different characteristics of the input. Phonemes are provided to the system and it recognizes them on the basis of their membership values under the defined fuzzy rule base. The experiments are performed on data from 180 speakers out of which 130 samples are used for training while the remaining 50 are used for testing.