Fundamental Frequency Contour Synthesis for Turkish Text-to-speech

FUNDAMENTAL FREQUENCY CONTOUR SYNTHESIS FOR TURKISH TEXT-TO-SPEECH The subject of this M.S. thesis is fundamental frequency contour synthesis of Turkish sentences for text-to-speech systems. As intonation and stress are the most outstanding determinants of fundamental frequency, the work mainly focuses on these two prosodic phenomena. A detailed analysis was performed on the fundamental frequency contours of the sentences and the findings were presented. Combining the findings with the theoretical background, a fundamental frequency contour synthesis system is devised, relying on a sentence database and utilizing the syntactic structure of sentences based on word categories and word stress information. The work is based on the recordings of a single speaker. Different from the conventional prosody generation systems, prosodic phrasing and accent labeling, requiring an extensive text analysis, are not included in the system. The method yielded encouraging results for future research.