Research Of The Speech Recognition Technology Based On DSP

Voice is the exchange between people of the most direct way. Speechrecognition is an important means of human-computer interaction, which in manyindustries have been widely used. Recent years with the rapid development ofembedded technology, based on DSP voice recognition system voice recognition isbecoming an important development direction.This paper studies the technology and speech recognition algorithm is proposedto DSP chip as the core components to the specific DTW algorithm as the core ofpeople isolated word speech recognition system.The hardware part of the TMS320VC5402 DSP chip as the core architecture,including the audio capture module, power module, watchdog circuit, externalmemory modules, data acquisition and transmission modules and CPLD logiccontrol module and several modules. Acquisition chip TLV320AIC23B by voicevoice signal acquisition and conversion, and converted the signal sent to the DSP'smemory for processing, the recognition results obtained. System software from thespeech signal acquisition, data preprocessing, sub-frame windowing, featureextraction, DTW algorithm for endpoint detection and implementation of severalcomponents. First with the MATLAB software to simulate the preparation of thealgorithm, then C/C++ language and assembly language programming mix, andtransplant program results to the CCS platform for debugging. In this paper,detection algorithm and DTW algorithm was improved and optimized. The testproved that the improved algorithm in the recognition rate of loss of the system isnot the case effectively improve the recognition rate, so that the system hasimproved the overall recognition performance.