Research and application of audio feature in compressed domain

Author

Liaoyu Chang ; Xiaoqing Yu ; Haiying Tan ; Wanggen Wan

Author_Institution

Sch. of Commun. & Inf. Eng., Shanghai Univ., Shanghai

fYear

2007

fDate

12-14 Dec. 2007

Firstpage

390

Lastpage

393

Abstract

In this paper, by analyzing audio features in compressed domain based on audio encoding/decoding theory, we investigate the feature extraction directly from MP3 (MPEGl-layer3) compressed data stream and propose how to calculate these features such as RMS (root mean squared), SC (spectral centroid), BER (band energy ratio), BW (band width) and MFCC (Mel-frequency cepstral coefficients) from the spectral information available in the decoding stage. Also, the experiments are conducted and the results are analyzed to show the application of some aforementioned features. All the work conducted is for the purpose of laying a foundation for realizing audio information classification, retrieval and recognition in MP3 audio format.

Keywords

audio coding; data compression; decoding; feature extraction; information retrieval; mean square error methods; MP3; MPEGl-layer3; Mel-frequency cepstral coefficients; audio encoding-decoding theory; audio feature; audio information retrieval; audio recognition; band energy ratio; domain compressibility; feature extraction; information classification; root mean square; spectral centroid; MFCC; audio feature; compressed domain; encoding/decoding;

fLanguage

English

Publisher

iet

Conference_Titel

Wireless, Mobile and Sensor Networks, 2007. (CCWMSN07). IET Conference on

Conference_Location

Shanghai

ISSN

0537-9989

Print_ISBN

978-0-86341-836-5

Type

conf

Filename

4786220