| Summary: | 碩士 === 長庚大學 === 資訊工程學系 === 104 === In this study,we use the cloud TTS (Text-to-speech)combined with speech recognition (Speech Recognition) ,to designed a convenient sound-text sync audiobooks systems. Users can use their favorite articles to make their own learning materials in Chinese,English, and Japanese. For different languages we do the hyphenation to split article to sentences, pinyin conversion, producing Pinyin sentence, with the time point form the cloud TTS. Processed by the CGUAlign which is using Python to package the HTK(Hidden Markov Model Toolkit) to make the word-level time text file. And search the location for all the sentences and words in article. Final use sentences,words, time points and locations to make the tlt(text-location-time) file to save the article. By testing for 100 English articles this system can reach 86% of success rate to make the sound-text sync audiobooks.
Users only provide a text file by using the system design from Python tkinter moduel. Users can make their own learn materials and save it. User can choose any sentence or word in article to play it`s voice repeatly,for improving their listening and speaking ability.
Source code: https://github.com/CHANGCHAOKAI/Google_tts-book
|
|---|
