Speaker Recognition for Digital Forensic Audio Analysis using Support Vector Machine

Abstract

Speaker Recognition is included in pattern recognition, where one of the most critical parts is the process of data classification. In the classification, the built system must estimate the classification of data into a classroom dimension closest to the training set. The speaker's introduction aims to identify evidence of speech recording by a handheld telephone that involves comparing one or more unidentified sound samples with one or more known sound samples. In this research, the data used in the form of evidence of recording conversation by telephone and recording of comparison of some unexpected. The part that is done is to classify speaker recognition with the Support Vector Machine (SVM) classification method to recognize the speaker. Using the SVM method, the accuracy of classifying the speaker's introduction is excellent. From the test results, the SVM method's use resulted in an accuracy rate of 86.67% for the test with the same sentence and up to 67% for different sentences to recognize the speaker with the values of C 0.01 and $\boldsymbol{\gamma}$ (Gamma) 0.0001.