Audio stream analysis for environmental sound classification

We present in this paper a framework for audio concept identification based on audio stream analysis and binary classifiers encapsulation. The system consists of three stages. The first stage is called the pre-processing level audio, where audio stream is segmented and silence segments are detected. In the second stage, speech, music and environmental sounds are automatically divided and further classified. This classification is based on time frequency analysis of audio signals. In the third stage, a novel framework encapsulating of binary classifiers is implemented. It is shown that the proposed system has achieved accuracy higher than 90% for audio concept identification.