International Journal of Mechanical Engineering and Research

Open Access Open Access  Restricted Access Subscription Access
Open Access Open Access Open Access  Restricted Access Restricted Access Subscription Access

A Novel System for Speech Recognition


Affiliations
1 Sri Vani School of Engineering, Vijayawada
2 JNTU College of Engineering, Kakinada
3 SRKR Engineering College, Bhimavaram
     

   Subscribe/Renew Journal


Neural networks are excellent classifiers, but performance is dependent on the quality and quantity of training samples presented to the network. In cases where training data is sparse or not fully representative of the range of values possible, incorporation of fuzzy techniques improves performance. That is, introducing fuzzy techniques allow the classification of imprecise data. The neuro-fuzzy system presented in this study is a neural network that processes fuzzy numbers. By incorporating this attribute, the system acquires the capacity to correctly classify imprecise input.

Experimental results show that the neuro-fuzzy system's performance is vastly improved over a standard neural network for speaker-independent speech recognition. Speaker independent speech recognition is a particularly difficult classification problem, due to differences in voice frequency (amongst speakers) and variations in pronunciation. The network developed in this study has an improvement of 45% over the original multi-layer perceptron used in a previous study.


User
Notifications
Font Size


  • Kosko, Bart, Neural Networks and Fuzzy Systems: A Dynamical Approach to Machine Intelligence, Prentice Hall, Englewood Cliffs, NJ, 1992.

  • Nava, P. and J. Taylor, “The Optimization of Neural Network Performance through Incor-poration of Fuzzy Theory,” Proceedings of The Eleventh International Conference on Systems Engineering, pp. 897-901.

  • Alefeld, G. and J. Herzberger, Introduction to Interval Computation, Academic Press, New York, NY, 1983.

  • Bojadziev, G. and M. Bojadziev, Fuzzy Sets, Fuzzy Logic, Applications, World Scientific, New Jersey, 1995.

  • Ishibuchi, H., R. Fujioka, and H. Tanaka, “Neural Networks That Learn from Fuzzy If- Then Rules,” Vol.1, No.2, pp. 85-97, 1993.

  • Nava, P., New Paradigms for Fuzzy Neural Networks, Ph.D. Dissertation, New Mexico State University, Las Cruces, New Mexico, 1995.

  • Carnegie Mellon University Neural Network Benchmark Collection, Pittsburgh, PA, 1994.

  • Deterding, D.H. Speaker Normalization for Automatic Speech Recognition, Ph.D. Dissertation, Cambridge, England, 1989.

  • Robinson, A.J., Dynamic error Propagation Networks, Ph.D. Dissertation, Cambridge

  • University, Cambridge, England, 1989.

  • Grana, M., A. D’Anjou, F. Albizuri, J. Lozano, P. Laranaga, M. Hernandez, F. Torrealdea, A. de la Hera, and I. Garcia, “Experiments of fast learning with High Order Boltzmann Machines,” Informatica y Automatica, 1995.

  • Nava, P. and J. Taylor, “Voice Recognition with a Fuzzy Neural Network,” Proceedings of The Fifth IEEE International Conference on Fuzzy Systems, pp. 2049-2052.

  • Zadeh, L.A., Fuzzy Sets and Applications: Selected Papers by L. A. Zadeh, John Wiley & Sons, New York, 1987.


Abstract Views: 478

PDF Views: 6




  • A Novel System for Speech Recognition

Abstract Views: 478  |  PDF Views: 6

Authors

Ande Stanly Kumar
Sri Vani School of Engineering, Vijayawada
K. Mallikarjuna Rao
JNTU College of Engineering, Kakinada
A. Bala Krishna
SRKR Engineering College, Bhimavaram

Abstract


Neural networks are excellent classifiers, but performance is dependent on the quality and quantity of training samples presented to the network. In cases where training data is sparse or not fully representative of the range of values possible, incorporation of fuzzy techniques improves performance. That is, introducing fuzzy techniques allow the classification of imprecise data. The neuro-fuzzy system presented in this study is a neural network that processes fuzzy numbers. By incorporating this attribute, the system acquires the capacity to correctly classify imprecise input.

Experimental results show that the neuro-fuzzy system's performance is vastly improved over a standard neural network for speaker-independent speech recognition. Speaker independent speech recognition is a particularly difficult classification problem, due to differences in voice frequency (amongst speakers) and variations in pronunciation. The network developed in this study has an improvement of 45% over the original multi-layer perceptron used in a previous study.


References