Speech Recognition Design And Optimize Research In Embedded System

Speech recognition is a kind of technology that enables device to recognize and comprehend speech signal. It processes speech signal, involving linguistics and computer science, signal processing, psychology, etc. This technique is recognized as the most challenging and has high practical value. It has very important application prospect on Internet, communication, national defense and the people's livelihood field.However the speech recognition implementation in embedded systems is influenced by many factors. In embedded systems environment, computing rate is relatively lower, storage space is limited, and requires certain real-time. Although the speech recognition technology won some achievement in high performance platform, it's necessary to do further analysis and research on speech recognition in embedded platform.This paper's main research works are listed as follows:1) Constructing speech recognition hardware system. Select high performance processor as the key processing unit, and design outlying work module.2) Construct speech recognition software system, then design and transplant the embedded Linux system.3) The system uses calculation of linear predictive coding MEL cestrum coefficients as the characteristic parameters of speech samples. Use hidden markov model of high recognition rate to identify recognition problem.4) Based on the actual system condition, optimize the input signal, hardware working mode and speech recognition algorithm.5) After testing validation, this system is of high recognition speed and accurate recognition rate, realizes the speech recognition function at the same time.