1.Algorithm of Abnormal Audio Recognition Based on Improved MFCC 2012.pdf

Procedia Engineering 29 (2012) 731 – 737 1877-7058 ? 2011 Published by Elsevier Ltd. doi:10.1016/eng.2012.01.032 Available online at Available online at Procedia Engineering 00 (2012) 000–000 Procedia Engineering /locate/procedia 2012 International Workshop on Information and Electronics Engineering (IWIEE) Algorithm of Abnormal Audio Recognition Based on Improved MFCC Chuan Xiea,Xiaoli Caoa,Lingling Hea aCollege of computer science and information engineering, Chongqing technology and business university, Chongqing 400067, china Abstract Characteristics extraction has a great effect on the audio training and recognition in the audio recognition system. MFCC algorithm is a typical characteristics extraction method with stable performance and high recognition rate. For the situation that MFCC has a large amount of computation, an improved algorithm MFCC_E is introduced. The computation of MFCC_E is reduced by 50% compared with the standard algorithm MFCC, and it make the hardware implementation is easy. The experimental result indicated that MFCC_E and MFCC have the same recognition rate roughly, yet the computational complexity of MFCC_E is much smaller. ? 2012 Published by Elsevier Ltd. Keywords：audio recognition; characteristics extraction; MFCC; MFCC_E; GMM 1. Introduction In the field of target tracking, the video and the radio data are the two most important kinds of information, and in the past two decades, the video tracking has been in a dominant position, but its tracking performance will be greatly reduced when the tracking target is out of the observation range. Compared with video tracking, acoustic sensor has the advantages of low cost, small size, high efficiency, and the audio signal changes slowly over time, so the collected audio signal is stable and reliable. Therefore, the audio recognition and the audio target localization have become the research hotspots in recent years [1] . In the field of au