Synthesis by Rule of Some Retroflex Speech Sounds

Abstract

This paper describes a formulation of rules for the synthesis of Tamil retroflex sounds. Based on palatograms, the sounds /t, n, / were determined to be palatal retroflex sounds and based on spectrograms in V—V context the relevant acoustic features were extracted and using them as input to a computer programme, these sounds were synthesized. A computer based SPEECH programme was used to synthesize these sounds and rules were formulated. The results obtained confirm the importance of the third formant and the transitions of vowels in the synthesis of Tamil retroflex sounds. This further confirms the validity of the structural classification of speech sounds on the basis of place and manner of articulation mentioned in Rao and Thosar (1968a, b).