Gaussian Filter-Based Speech Segmentation Algorithm for Gujarati Language

Abstract

AbstractAutomatic speech segmentation is a main step in speech signal production and analysis process. Great advancement in speech synthesis has already been made using concatenative algorithms. Syllable is most suitable speech unit for concatenative speech synthesis because it does not require extensive prosodic models and provide better co-articulations than other sound units. To get natural sounding, output speech segmentation plays very important role. Speech segmentation is the process of dividing speech signal in to smaller units of sound. So accurate selection of speech unit and detection of boundaries are very important. In this research work, Gujarati language is used for segmentation and database is created. This paper suggests a method of syllable segmentation to detect boundaries of syllable by means of start point of syllable and end point of syllable. Performance parameters such as accuracy and peak signal to noise ratio (PSNR) are evaluated. Producing natural sounding speech signal in different Indian languages is a very demanding and ongoing problem.KeywordsText to speech (TTS)SyllableGaussian filterStart point (SP)End point (EP)Segmentation