Robust noise MKMFCC–SVM automatic speaker identification

This paper proposes robust noise automatic speaker identification (ASI) scheme named MKMFCC–SVM. It based on the Multiple Kernel Weighted Mel Frequency Cepstral Coefficient (MKMFCC) and support vector machine (SVM). Firstly, the MKMFCC is employed for extracting features from degraded audio and it uses multiple kernels such as the exponential and tangential and for MFCC’s weighting. Secondly, the extracted features are then categorized with the SVM classification technique. A comparative study is performed between the proposed MKMFCC–SVM and the MFCC–SVM ASI schemes using the MKMFCC and MFCCs with five schemes for extracting features from telephone-analogous and noisy-like degraded audio signals. Experimental tests prove that the proposed MKMFCC–SVM ASI scheme yields higher identification rate in noise presence or degradation.