Abstract: This paper investigates the significance of Supra-Segmental features in continuous speech of Tamil. The Supra-Segmental parameters like fundamental frequency, pitch, intensity, time duration, rhythm and etc., can help to signal the syntactic structure of utterances into larger discourse segments and provide additional information for human speech processing. It also provides information about the linguistic structure of speaker’s message and the emotional state. In human speech processing, linguistic context and phonological rules help the brain to separate syntactic units into phonemes, syllables, words, sentences and phrases. In the case of agglutinative languages, searching space reduction is very important during automatic speech processing. The entitled study focused on Tamil language, which belongs to the family of languages called Dravidian and noted for its highly agglutinative nature. Here the aim is to examine and prove that prosodic information carried out acoustically by the speech signal can be used to improve the performance of speech processing and to add syntactic, semantic level functionality to it. Therefore, the study introduces a rule-based model which shows the relationship between the Supra-Segmental parameters of phoneme to sentence level and their statistics. Finally, the results were compared to analyse the accuracy and efficiency of the model. The methods introduced here are easily adoptable to other agglutinative languages. Instead of using the prosodic level boundaries the study make use of statistical properties, which are more advanced.

Keywords: Tamil, Supra-Segmental, Fundamental frequency, Duration, Intensity, PRAAT.