Research on Speech Recognition Realization Based on Deep Learning Method
摘要
Automatic speech recognition is an important research direction in artificial intelligence field. For how to implement speech recognition, converting speech sequences into text sequences, it is simply a matter of identifying the problem, selecting a model and then training it. With the continuous expansion of the open source community, the research process in the field of speech recognition has been accelerated. Some open source tools for speech recognition, such as CMUSphinx, Julius, HTK, CMUSphinx and ISIP, have also emerged and been widely used by researchers. Firstly, the paper analyzes the tools CMUSphinx, Kaldi and deep learning platform which can develop speech recognition at present. Then the implementation process of CMUSphinx open source tool is studied. Secondly, the practice process of speech recognition by Kaldi is discussed. Finally, some key problems in speech recognition implementation and future research directions are pointed out in the paper.