Abstract
Aiming at the problem that the performance of speech recognition system will drop severely in noisy environments, this paper proposed a recognition system that has excellent anti-noise performance. The feature parameters of front-end are ZCPA (Zero-Crossings with Peak-Amplitudes) feature and the recognition network of back-end is wavelet neural network. Mexican Hat wavelet was used to replace the Sigmoid or Gaussian basis function of feed-forward neural network. The network structure was three layers. Training network weights, determining the position and scale factor of wavelet function were processed respectively. And the information of training samples was made full use of when estimating the position factor. Hence, the network could converge rapidly and it also avoided the problem of wavelet “dimension disaster”. A 50 isolated-word person-independent speech recognition system using BP network or wavelet neural network as recognition network was simulated under different SNRs in this experiment. The experimental results showed that recognition rates using wavelet neural network are higher than using BP network.
This work was partially funded by Shanxi Province Foundation for Returness(2009-31), International Scientific and Technological Cooperation Projects in Shanxi Province(2008-081026).
Access provided by Autonomous University of Puebla. Download to read the full chapter text
Chapter PDF
Similar content being viewed by others
References
Yong, X., Aina, Q.: Noise robust speech recognition based on improved hidden markov model and wavelet neural network. Computer Engineering and Applications 46, 162–164 (2010)
Zhang, X., Li, H.: A passwords recognition system based on zcpa feature parameter. Electronic Technology 7, 27–29 (2010)
Hou, X.: Noise-robust speech recognition based on wavelet network and RBF network. Computer Engineering and Applications 45, 150–152 (2009)
Steven, G., Pan, A., Seddeik, Y.M.: A feature based technique for face recognition using mexican hat wavelets. In: PACRIM, pp. 792–797 (2009)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2011 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Wang, Y., Zhao, Z. (2011). A Noise-Robust Speech Recognition System Based on Wavelet Neural Network. In: Deng, H., Miao, D., Lei, J., Wang, F.L. (eds) Artificial Intelligence and Computational Intelligence. AICI 2011. Lecture Notes in Computer Science(), vol 7004. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-23896-3_48
Download citation
DOI: https://doi.org/10.1007/978-3-642-23896-3_48
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-23895-6
Online ISBN: 978-3-642-23896-3
eBook Packages: Computer ScienceComputer Science (R0)