Prosodic Analysis And Modelling For Malay Emotional Speech Synthesis
Main Article Content
Abstract
This paper discusses an emotional prosody generator for a Malay speech synthesis system that can re-synthesize the selected vocal emotion from neutral synthesized speech output and improve the naturalness by adopting rule-based prosody conversion techniques. The role of prosodic features in emotional expression, particularly fundamental frequency and duration, has been widely investigated in several research projects. This project attempts to improve the naturalness of the synthesized emotional Malay speech by establishing an effective mechanism for the re-synthesis of emotion. Such a mechanism is created by analyzing the variation in the F0 contour of continuous emotional Malay speech against a fixed time period. The emotional prosodic generator for Malay developed in the course of this research makes use of principles of parametric prosody manipulation to synthesize four basic emotions, namely happiness, anger, sadness and fear. Subjective evaluation by means of listening tests was conducted to validate the ability of the emotions generator to generate the necessary prosody to synthesize emotional expression. The evaluation results show an overall recognition rate of between 61% and 85%.