Accelerate training by DWT in speaker identification using SVMs

Support vector machines have got more attention in speaker verification and identification recently. It can get superior performance in classifiers and require significantly less training data. But training a support vector machine is equivalent to solving a linearly constrained quadratic programming problem in a number of variables equal to the number of data points. This optimization problem is known to be challenging when the number of data points is exceeds few thousands. We present art efficient approach to accelerate the training process for support vector machines on the very large data set in speaker identification. The discrete wavelet transform has been adopted and the results show that the performance has been improved.