Robust Speech Recognition based on Dynamical Selections

L. Ge, K. Shirai (Japan), and Y. Ge (PRC)


artificial intelligence, speech recogntion, robustness 523-146 314


Robust recognition theory has become one of research fo cuses of acoustic speech recognition. Acoustic speech digi tal signal is a random process repeatedly alternating sta tionary pieces with non-stationary pieces. However current used characteristic parameters based on linear and station ary hypothesis drawn from such signals and the rigid rec ognition models do not adapt to such repeatedly alternating property of acoustic speech. Though Missing Feature Ap proach (MFA) has been proved a considerable solution of enhancement of robustness for noisy speech, MFA classify ing in binary way seems to be rough and it cannot used to deal with cepstral feature. Consequently, current noisy speech recognition systems perform mostly poorly. This paper tries to set up dynamic recognition theory that applies non-linear doubly random time series instead of linear model, and auto-select types of parameters and recognition models based on non-stationary measure of real-time. Meanwhile this theory gives two approaches of Feature with Confident Weight (FCW) in three means to describe the effect of noise in a more precise way and available in cepstral domain. Experimental results show proposed ap proaches could improve the recognition accuracy signifi cantly in adverse environment, including stationary and non-stationary noise environments.

Important Links:

Go Back