Summary: | 碩士 === 國立勤益科技大學 === 電機工程系 === 99 === In this paper, the Mandarin text-to-speech technique is employed to accomplish the implementation of voiced E-book on the Android platform. Since a transformation from the text of E-book to the corresponding speech, it makes the reading more effortless and relaxed. The text-to-speech(TTS) system, which consists of the text analysis, the recurrent neural network(RNN) based prosodic generator, the synthesis unit with 411 Chinese syllable waveforms, and the pitch-synchronous overlap and add (PSOLA) speech synthesizer, is the core of the system proposed and is implemented with C++ programming language. Then, the user interface of the system proposed is achieved by using the Java language. Finally, the experimental results confirm that the complete system with memory required of 5MB is able to provide the fluent and clear speech in real-time.
|