With the AELMS filter, which can preserve sources features of speech and decrease the damage on speech information, noise of a contaminated speech signal got canceled, and a gaussian model was clustered as a method to make noise more robust. By composing a gaussian clustering model, which is a robust speech recognition clustering model, in a noise environment, a recognition performance was evaluated. The study shows that SNR of speech, which was gained by canceling the environment noise which was kept changing, was enhanced by 2.7dB in an average and a recognition rate was improved by 3.1%.
展开▼