Automatic speech recognition using Support Vector Machine and Particle Swarm Optimization
Gracieth C. Batista, Washington Luis Santos Silva, Angelo Garangau Menezes · 2016
Support Vector Machine (SVM) is an algorithm that trains and classifies different types of data through of an optimal hyperplane of decision. On the other hand, Particle Swarm Optimization (PSO) is, in general, an algorithm that finds the best point to represent a dataset. In this paper, PSO is used to find the best data of each class (pattern) to be trained by SVM and there is a comparison of the difference between using or not this optimization. The digits of zero to nine in Brazilian Portuguese language are recognized automatically by SVM. Those digits are pre-processed using melcepstral coefficients and Discrete Cosine Transform (DCT) to generate a two-dimensional matrix used as input to the PSO algorithm for generating the optimal data.