期刊名称：International Journal of Computer Trends and Technology
电子版ISSN：2231-2803
出版年度：2019
卷号：67
期号：4
页码：79-84
DOI：10.14445/22312803/IJCTT-V67I4P118
出版社：Seventh Sense Research Group
摘要：Voice processing is emerging as an incredible means of human communication. Increased dependency on machines by humans led to the development of voice recognition, mainly for human machine communication. It has become an inevitable part of medical, military and other security applications in recent years. Specific parameters that differentiate one speaker from another are the features, their extraction and classification describes every voice recognition system. Several voice recognition algorithms are evolved under various environmental conditions. MKMFCC (Multiple Kernel Weighted Mel Frequency Cepstral Coefficients), as a feature extraction algorithm is providing better feature extraction even in noisy or degraded environment compared to other existing algorithms. Besides, VQ (Vector Quantization), because of its low computational burden is a good classifier. This paper proposes a voice recognition system combining the good qualities of MKMFCC and VQ, with MKMFCC as the feature extraction algorithm and VQ as the classifier. The proposed system provides an accuracy of about 92.6% for the voice recognition process.
