Support Vector Machine Classifier for Pattern Recognition

Abstract

Automatiuc speech recognition is carried out by Mel-frequency cepstral coefficient (MFCC). Linearly-spaced at low and logarithmic-spaced filters at higher frequencies are used to capture the characteristics of speech. Multi-layer perceptrons (MLP) approximate continuous and non-linear functions. High dimensional patterns are not permitted due to eigen-decomposition in high dimensional image space and degeneration of scattering matrices in small size sample. Generalization, dimensionality reduction and maximizing the margins are controlled by minimizing weight vectors. Results show good pattern by SVM algorithm with Mercer kernel.