Automatic detection of Chinese accent-index based on approximation-ratio

Abstract

For a TTS system, to synthesize speech with better prosody, accent information is expected to be involved besides other. Therefore, we defined a set of Accent Index (AI) to represent the variances of accent in Chinese speech, and proposed a novel method to automatically annotate Chinese speech with AI. In the method, a parameter, named Approximation-Ratio, was used to numerically indicate the accent of prosodie unit. And the value of AI was the discretization of Approximation-Ratio. One corpus was annotated with AI by the method. And with the corpus, a refined prosody parameter prediction model was built. The experiment results showed that prosody parameters predicted by the refined model were more close to ones of real speech than by the former model without AI. Further, a perceptual evaluation showed that the accent manifestation generated by AI-ready synthesizer was distinguishable and acceptable. © 2004 IEEE.