Design Of Speech Recognition Module Of Smart Remote Controller

With the rapid development of the smart home industry, the technology of speech recognition is potentially of broader interest for applications in which the utilization has become a major concern in intelligent remote control to realize home control in the field of the smart home. Therefore, the design of a practical available embedded speech recognition system is of great importance.This thesis aims to design an embedded speech recognition module that has characteristics of low cost, easy operation and high recognition rate and could be applied to the intelligent remote control using the chip SEP6110C, an embedded chip. After comparing the related technology of embedded speech recognition, vector quantization is utilized in the thesis as the core algorithm for Pattern Recognition to complete the design of algorithms based on SEP6110C embedded platform. Meanwhile, in the process of voice preprocessing, simplify the design of filter and cancel framing. In the process of voice activity detection, devise determined rules and thresholds. When extracting the voice feature, weight MFCC parameters, so as to make it better show the feature of the voice. And finally utilize LBG algorithm for vector quantization.After the completion of the design, the function of the system is verified in the related environment. The results show that, the number of voice commands that the designed system can recognize is 30, and the speech recognition rate is more than 90%, and the time of the reaction is less than 1 second. Therefore, the function of speech recognition of a small vocabulary and isolated word is realized and requirements of home control in intelligent remote control are met in the system. So the design has an extensive application value in smart home field.