DocumentCode :
2462068
Title :
Pornographic Audios Detection Using MFCC Features and Vector Quantization
Author :
Qu, Zhiyi ; Yu, Jing ; Niu, Qiang
Author_Institution :
Sch. of Inf. Sci. & Eng., Lanzhou Univ., Lanzhou, China
fYear :
2010
fDate :
17-19 Dec. 2010
Firstpage :
924
Lastpage :
927
Abstract :
Identifying the phonetic characteristics of speakers is an important branch of speech recognition. The audition system of human being is an ideal speaker recognition system. MFCC (Mel-frequency cepstral coefficients) characterizes the auditory features of humans effectively, and thus has been widely used in practice. This paper explores applying MFCC and VQ (vector quantization) algorithms on the pornographic audios detection. Firstly, MFCC of selected pornographic audios are extracted and then encoded into codebooks using VQ algorithm, Secondly, all of the codebooks obtained will be averaged to get an average codebook, Finally, the type of any newly input audio belonging to, either pornographic or non-pornographic, will be determined by measuring the Euclidean distance between the average codebook and its own codebook. Experiment results show that the algorithm can detect pornographic audios effectively.
Keywords :
audio signal processing; cepstral analysis; speaker recognition; speech processing; vector quantisation; Euclidean distance; MFCC features; audition system; auditory features; codebook encoding; mel-frequency cepstral coefficients; phonetic characteristics; pornographic audios detection; speaker recognition system; speech recognition; vector quantization; Euclidean distance; Feature extraction; Filter bank; Humans; Mel frequency cepstral coefficient; Training; Vector quantization; Euclidean distance; Mel-frequency cepstral coefficients; pornographic audios detection; vector quantization;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computational and Information Sciences (ICCIS), 2010 International Conference on
Conference_Location :
Chengdu
Print_ISBN :
978-1-4244-8814-8
Electronic_ISBN :
978-0-7695-4270-6
Type :
conf
DOI :
10.1109/ICCIS.2010.228
Filename :
5709240
Link To Document :
بازگشت