Studies On Techniques For Chinese Speech Simulation System Based On Corpus

This thesis discusses on speech simulation technology which is based on the advanced corpus-based speech synthesis. And our research focuses on the manufacture of speech simulation system. Considering the main problems of traditional Mandarin Text-to-Speech System, in-depth research was conducted on a series of key techniques such as text prosodic level marking, corpus analysis and design, unit selection strategy and etc. We firstly take a glance back at the history of Mandarin speech synthesis technology whose defects is also indicated. It is important to manufacture the speech simulation system, so the generation of speech and the principium of speech synthesis are introduced. And the characteristic of Text-to-Speech(TTS) System is also shown.Then we introduce the text prosodic level marking in detail, analyze the prosodic level structure of sentences, point out the pause principle in sentences and the description of lexical word. We use C4.5 algorithm to label text prosodic levels automatically. Thirdly we discuss upon corpus analysis and design and determine how to select speech synthesis units.In the design of corpus, we carefully analyze the syllable distribution of corpus TH-CoSS, then classify the prosodic characters of this corpus and present out the distribution of every prosodic character. Based on prosodic character vector, we construct an error function which is used to select original corpus for simulation system, and show the distribution of prosodic characters for the original corpus. Greedy algorithm and corpus self-adaptive process are expatiated to set theoretical foundation for text material search.For the unit selection strategy, we achieve candidate units for target ones by calculating target cost based on prosodic character vector. Then we use Viterbi algorithm to select the best synthesis path for simulation speech by calculating concatenative cost of the synthetic waveform, thus a corpus-based speech simulation system comes into being, and speech simulated by this system has the original speaker's style, rather vivid.