Data Mining and Knowledge Engineering

Shrinivas P. Mahajan ¹, Jyotsana Sahu ², Mukul S. Sutaone ³, V. K. Kokate ²

Affiliations
1 Department of Electronics and Telecomm, College of Engineering, Pune, IN
2 College of Engineering, Pune, IN
3 Electronics and Telecommunication Department, IN

Abstract

Audio classification has found widespread use in many emerging applications. It involves extraction of vital temporal, spectral and statistical features, and using these in creating an efficient classifier. Most of the audio classification work has been done on binary class classification. In our work we suggest best suited features for classification of different audio classes. Here, we present an algorithm for audio classification that is capable of segmenting and classifying an audio stream into speech male, speech female, music, noise and silence. The speech clips are further segment into voiced and unvoiced frames. A number of timbre features have been discussed, which distinguish the different audio formats. For pre classification, Probability Density Function (PDF), which is a threshold-based method, is performed over each audio clip. For further classification, K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) Classifiers are proposed. Experiments have been performed to determine the best features of each binary class. Utilization of these features in multiclass classification yielded accuracy 96.34% in audio discrimination.

Keywords

Audio Feature Extraction, Bayesian Classification, K-Nearest Neighbor, Support Vector Machine.

Full Text

     

Shrinivas P. Mahajan ¹, Aparna S. Joshi ²

Affiliations
1 Department of Electronics and Telecomm., College of Engineering, Pune, IN
2 Vishay Inter Technologies Ltd., Pune, IN

Abstract

Automatic audio classification is an important part in audio indexing, content-based audio retrieval and online audio distribution. The efficiency of an audio classification or categorization depends on the ability to capture proper audio features and to accurately classify each feature set corresponding to its own class. In this paper a novel hierarchical audio classification algorithm using Support Vector Machine(SVM) is proposed to classify the  input digital audio stream into five audio classes namely male speech,female speech, music, silence and environmental sound. An audio database “Swarsanchay” consists of recordings of 1000 audio clips of all the above mentioned audio classes with different noisy conditions was created for experimentation. Out of which five hundred clips were used for training and remaining five hundred audio clips were used for testing. The proposed multi class audio classifier correctly identified 475 audio clips giving an overall classification accuracy of 95%. During our investigation of feature extraction we identified best ranking features pertinent to class having better discriminatory power amongst the classes. The best classification accuracy of 95% is obtained when best ranking feature set is used for hierarchical multiclass classification.

Keywords

Audio Classification, Audio Features, Feature Extraction, Support Vector Machines.

Full Text