Comparative Performance Analysis of SVM Speaker Verification System using Confusion Matrix

In Speaker verification task, it is necessary to calculate the performance of the speaker verification system; there are many systems available for the speaker verification task which uses the different type of modeling schemes like generative modeling and discriminative modeling. We are using discriminative modeling with the help of Support vector machine for speaker verification task. We propose the use of confusion matrix for the performance calculation of support vector machine in contrast with the parameters like accuracy and precision. Speaker verification includes training of feature vector and then the classification of trained feature vector. Every kernel of support vector machine gives the different performance for the classification task so with the help of this confusion matrix approach we can also compare the different kernel performances. The parameters of performance make us capable for selecting the best kernel for our data sets.