A robust tamil text to speech synthesizer using support vector machine (svm)

Abstract

Speech synthesis systems aim at generating high-quality, natural-sounding speech. Synthesizers become more static due to the inaccessibility of large databases for Indian languages. The prominence of the artificial speech made by these synthesizers is poor. Though statistical-based technique on hidden Markov models (HMMs) and Gaussian mixture model (GMM) is a powerful technique in Text to Speech (TTs) synthesis, recent work in TTS has concentrated on support vector machine (SVM). In this paper, a TTS system is developed for Tamil language using SVM. By using SVM, better sensitivity measures for Tamil Text to Speech are obtained. Output portrays that SVM can produce effectively natural speech compared to HMM and GMM.