Thesis Details
Záznam streamovaného audia
Speech group at the Faculty of Information Technology have very good results in the language identification. For further improvements in this field it is necessary to get more data for training and testing identification tools. The main object of this project was to download streams from internet radios and to recognize speech blocks in received data. The first objective was to download stream in different languages, the second was to mark speech segments in the saved audio files using phonem recogniser and application ngram according to language models of speech and music. The project is trying to get best results using software avaible on our faculty and opensource applications.
downloading audio streams, segmentation, phnrec, ngram