Environmental Audio Classification Based on Active Learning with SVM

DOI

10.2991/csic-15.2015.50How to use a DOI?

Keywords

Active learning, Environmental audio classification, Support vector machines, SVM_EPS, MOA

Abstract

In order to solve the classification of environmental audio data under the fewer number of the training examples, this paper combined Support Vector Machines SVM) and Entropy Priority Sampling (EPS), and proposed the SVM_EPS method as the selecting sampling strategies in active learning. And the method MOA (Multi-variant Optimization Algorithm) was exploited to select the optimal kernel parameters of SVM. In experiments, the CELP features in 11 dimensions were extracted from the given environmental audio data, and the classification performances were compared under different percent training samples with SVM, EPS and SVM_EPS. The results show that SVM_EPS method outperforms the SVM and EPS.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Yan Zhang
AU  - Danjv Lv
AU  - Ying Lin
PY  - 2015/07
DA  - 2015/07
TI  - Environmental Audio Classification Based on Active Learning with SVM
BT  - Proceedings of the 2015 International Conference on Computer Science and Intelligent Communication
PB  - Atlantis Press
SP  - 208
EP  - 212
SN  - 2352-538X
UR  - https://doi.org/10.2991/csic-15.2015.50
DO  - 10.2991/csic-15.2015.50
ID  - Zhang2015/07
ER  -