Local Tetra Pattern Texture Features for Environmental Sound Event Classification

Abstract

Audio feature extraction and classification are important tool for audio signal analysis in many applications, such as home care system, security surveillance, meeting room sounds and music classification and so on. This paper presents sound classification by combining of image processing and signal processing to classify the data accurately. Firstly, audio signal is converted into time-frequency representation same as texture image in image processing. And then local tetra pattern (LTrP) text feature is used to extract features from this image. Finally, audio signal is classified by using one-vs-one SVM classifiers. Evaluation is tested on ESC-10 dataset.