Research Of Chinese Speech Synthesis Technology Based On Speech Database

Speech synthesis, also known as Text-to-speech(TTS),is a technique for generating intelligible artificial speech for a given input text. In recent years, with the rapid development of computer technology, speech synthesis has been used widely in various applications including in-car navigation systems, e-book readers, voice-over functions for the visually impaired, and communication aids for the speech impaired. As an important part of the human-computer interaction, speech synthesis will play a more and more important role in the people’s life in the future. Therefore, we studied the current technology of speech synthesis all over the world in this paper in detail, and on this basis, we build up an Chinese speech synthesis system.In This paper, we firstly introduce the concept of speech synthesis, the method to implement TTS and its history. Finally, we choose the method that based on hidden markov model(HMM) for our system. Then, we introduce the principle of HMM and its algorithm including forward-backward algorithm and Viterbi algorithm and Baum Welch algorithm. And we study the speech synthesis tools HTS,which is based on HMM. At last, the framework of this TTS system is determined.According to the characteristics of the Chinese grammar, we chose Chinese phonemes as element to build HMM model of speech acoustic parameters.In order to well noted the context information, we divided the Chinese sentence structure into six levels and design more than 200 classification problem sets for clustering based on decision-tree, and designed an Chinese speech database,which contains 550 audio files and accurate annotation text.Through HMM training by the parameter adjustment and optimization,we choose the final HMM model, and use the voice synthesizer to synthesize the Chinese speech data with intelligibility and naturalness.Finally, through the analysis of test results,we are clear about the speech synthesis system improvement direction in the future.