Computer-aided animated gesture-driven facial model with speech synthesis

Abstract

The aim of this study was to develop a gesture-driven facial model with speech synthesis capability. A two-dimensional facial Model was developed and animated based on the Facial Action Coding System. Such emotions as "happy", "sad", "anger", and "fear" were simulated and visualized through the combination of eight action units. A speech synthesizer for the Tamil language was built using a syllable-based concatenation approach. The results indicated that the synthetic speech had an average accuracy rate ranging from 85% to 90% as natural as the human speech. Moreover, 75%-85% of the words were articulated well and identified by the children correctly. The ultimate goal of the system is to assist children with vocal and hearing disabilities in their language learning process.