Research And Implementation On Speaker-independent Chinese Continuous Digit Speech Recognition System

After the study of speech recognition theories,using the Visual C++programmingtechnology,this article build a real-time speech recognition software on which we makea series of experiments and get good results.Firstly, this article introduces the state of the speech recognition techniques andsome difficultices in the implementation of Chinese continuous digit speech recognitionsystem, further gives related background materials and the purpose of this research.Secondly, the principles for the modeling and construction of Speech Recognitionsystem are explained, the speech signal analysis techniques and the theory of HiddenMarkov Model (HMM) are detailedly introduced. These laid the theoretical foundationfor implementation of speech recognition software and carrying out the experiments.Thirdly, we build a real-time speech recognition software system with theapplication of VC++software programming technology.In the process of building thesoftware systems, we take a full account of the software flexibility and ease of use. Inrecognition processing,we in-depth study and learn from the Cambridge UniversityHTK speech recognition system’s source code, to ensure that the advanced nature of therecognition system.Finally, we make a series of experiments of Chinese independent continuous digitalspeech recognition on the software. The related results are analyzed and identified,weachieve good results. At the end of this article, the comprehensive summary and theprospect of future work are also given.