Application Research On Speech Recognition In A Virtual Campus System

Using 'virtual campus roaming system of HuaZhong Normal University' as an object, this paper researches on problems about speech recognition techniques. A word-isolated, speaker-dependent, and small glossary speech recognition system is designed and implemented based on MATLAB and VC++ to combine voice navigation capabilities with virtual campus roaming system. It mainly includes three aspects: end checking, characteristic extraction and modeling matching. Original speech signal is collected and end checking is implemented through acoustics parameter calculation such as zero-crossing rate and energy. Using the coefficient of LPC of the signal in characteristic withdraw to compute its value is foundation of the speech recognition capabilities. Hidden Markov Model is adopted to trace and match the speech in modeling matching. Results show that HMM has a higher recognition rate and potential advantages in automatic speech recognition.In this paper, the Speech Application Program Interface in Speech SDK 5.1 is analyzed. The method of developing speech recognition application based on Speech SDK is summarized. Combined with HMM model, integrated with virtual campus roaming system, the voice navigation capabilities has been designed and implemented to make its operations more flexible, more convenient, and more humane.