Study Speech Recognition System Based on Manifold Learning

Abstract:

This paper conducts a comprehensive research and discussion on the relevant technologies and manifold learning.Traditional MFCC phonetic feature will lead a slower learning speed on account of it has high dimension and is large in data quantities. In order to solve this problem, we introduce a manifold learning, putting forward two new extraction methods of MFCC-Manifold phonetic feature. We can reduce dimensions by making use of ISOMAP algorithm which bases on the classical MDS (Multidimensional scaling). Introducing geodesic distance to replace the original European distance data will make twenty-four dimensional data, which using the traditional MFCC feature extraction down to ten dimensional data.

• Periodical: Applied Mechanics and Materials (Volumes 380-384)
• Pages: 3762-3765
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.380-384.3762
• Online since: August 2013
• Authors: Peng Hao Zhang
• Keywords: Feature Extraction, Geodesic Distance, Manifold Learning, MFCC-Manifold

References

[1] Bregler C., Omohundro S.M. ，Nonlinear Manifold Learning for Visual Speech Recognition, Int. Conf. Computer Vision，(1995).

DOI: 10.1109/iccv.1995.466899

[2] X. He and P. Niyogi，Locality Preserving Projections. Proc. 16th Conf. Neural Information Processing Systems，(2003).

[3] J．B．Tenenbaurn, Vin de Silva and J011ll C．Langford．A Global Geometric Framework for Nonlinear Dimensionality Reduction．Science．V01．290(5500)，PP．2319—2323，(2000).

[4] D．Liu，K．M．Lam and L．S．Shen．Illumination Invariant Face Recognition．Pattern Recognition．V01．38(10)，PP．1705—1716，(2005).

[5] Moghaddam B．Principal manifolds and Bayesian subspaces for visual recognition.