Speech recognition system based on visual features and genetic algorithm for persons with speechimpairments
Based on the movements of talkers’face,nose,mouth and throat are known to convey visual cues and represent several different kinds of information contained in the speech signal,and that can improve speech recognition rate,especially where there is noise or hearingimpairment.We proposed a new speech recognition method using these visual features and genetic algorithm.Genetic algorithm(GA) was first used to replace steepst descent method(SDM) and make a global search of optimal weight in neural network,the improved GA was then used to train the neural network.Six chinese vowels were taken as the experimental data,four handicapped speakers were taken as the testee.Recognition experiments show that the method is effective and high speed and accuracy for speech recognition.At present,the average recognition rate is 87.28％ at 10dB and 86.40％ at 0dB signal-to-noise ratio,So the features have very good robustness and antinoise,they are equivalent to an decrease of 8-10dB in the signal-to-noise ratio when speech sentences are presented in a noise background or are spoken by persons with hearing-impairment.
speech recognition visual feature genetic algorithm speech-impairment.
Zhi-yan Han Xu Wang Jian Wang
College of Information Science & Engineering,Northeastern University,Shengyang Liaoning,China
国际会议
International Conference on Modelling,Identification and Control(模拟、鉴定、控制国际会议)
上海
英文
2008-06-29（万方平台首次上网日期，不代表论文的发表时间）