Thai Expressive Speech Processing Technology: A Review

Problem statement: The studies on Thai expressive speech or emotional speech have been conducted for years. Most of them are expected to analysis the characteristics of Thai expressive speech. However, the conclusive reviews on these studies have not been conducted for further study on the speech technology or application of Thai expressive speech. Approach: The review of research on Thai expressive speech in various aspects has been performed. They include an analysis of fundamental frequency contours using Fujisaki’s model, an analysis of fundamental frequency contours using structural model and speech compression with noisy environments. It has been noted that four speaking emotions include enjoyable, sad, angry and reading styles. Results: A comparison of two successful F0 models has been reviewed. One approach is based on the Fujisaki’s model which has been applied for many tonal and toneless languages. Another one is based on the structural model which has been conducted primarily for Mandarin Chinese. Moreover, a study of speech compression for noise-corrupted Thai expressive speech by using two coding methods of CS-ACELP and MPCELP has been summarized. Conclusion: From the study, it can be seen that two mathematical models have been successfully applied to model the fundamental frequency contour of Thai expressive speech. As for speech compression, it can be seen that coding methods, types of noise, levels of noise, speech gender influence on the coding speech quality.