Abstract
Multi-view learning with multiple distinct feature sets is a rapid growing direction in machine learning with boosting the performance of supervised learning classification under the case of few labeled data. The paper proposes Multi-view Simple Disagreement Sampling (MV-SDS) and Multi-view Entropy Priority Sampling (MV-EPS) methods as the selecting samples strategies in active learning with multiple-view. For the given environmental sound data, the CELP features in 10 dimensions and the MFCC features in 13 dimensions are two views respectively. The experiments with a single view single classifier, SVML, MV-SDS and MV-EPS on the environmental sound extracted two of views, CELP & MFCC are carried out to illustrate the results of the proposed methods and their performances are compared under different percent training examples. The experimental results show that multi-view active learning can effectively improve the performance of classification for environmental sound data, and MV-EPS method outperforms the MV-SDS.
Highlights
Audio classification is an important access to extract audio structure and content, as well as a basis for audio retrieval and analysis
The classification of environmental audio requires a number of training examples that are too expensive or
The classification of environmental audio requires a number of training examples that are too expensive or tedious to acquire
Summary
Audio classification is an important access to extract audio structure and content, as well as a basis for audio retrieval and analysis. Semi-supervised learning and active learning, as methodologies of machine learning, make the best use of the unlabeled samples to assist the few labeled examples in establishing classifier model to improve the performance of classification even under the fewer number of the training examples. The experiments with the single classifier, EPS and SDS on the environmental sound are carried out in order to illustrate the results of the proposed methods and compare their performance under different percent training sample. The experimental results show that active learning can effectively improve the performance of environmental sound data classification, even under the fewer number of the training examples. The literature [9] combined support vector machines (SVM) and EPS, and presented the SVM_EPS methods as the selecting strategy in active learning for environmental audio data classification.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call