Multilingual speech recognition and speaker identification using soft computing

Abstract

In this present work an attempt is made to develop Multilingual speech and speaker newlinerecognition system based on speech features which are used to identity gender, newlinespeaker and recognize the speech and language among several speakers of known newlinespeech characteristics, from a sample of his or her voice. Every speaker has different newlineindividual characteristics embedded in his /her speech utterances. newlineTo evaluate speech characteristics from utterances they are stored in digitized form. newlineSpeech features namely linear prediction coefficients, linear prediction ancestral newlinecoefficients, zero crossing rates, mel-frequency ancestral coefficients, delta melfrequency newlinecepstral coefficients and formant frequencies are extracted from speech newlinesignal and speech feature vectors are formed. After that clustering was applied in the newlinespeech features. Then speech data features are fed into different models: Artificial newlineNeural Network (ANN) for training and testing processes of different speakers. We newlinesimulated ANN, which was trained by Back Propagation Algorithm (BPA) and Radial newlinebasis function (RBF). The best model has been adopted for getting the maximum newlineaccuracy for recognition of speech and speaker along with language and gender newlineidentification. The database used for this system consists of 50 speakers including newlineboth males and females from different parts of India and languages that have been are: newlineHindi, English, Sanskrit and Telugu.