Audio scene recognition method based on acoustic events

Abstract

The invention discloses an audio scene recognition method based on acoustic events, which comprises the steps as follows: Step I, conducting audio division on an audio stream to be subjected to audio scene recognition, Step II, classifying audio frames included in audio scene segments with an acoustic event model and obtaining a probability relationship among the audio frames and acoustic event classes, Step III, obtaining a probability relationship among the audio scene segments and the acoustic event classes according to the information of all the audio frames included in the audio scene segments, and Step IV, determining main acoustic events included in the audio scene segments and determining semantic scenes of the audio scene segments. The method judges the main acoustic events more reasonably and accurately, so that the recognition accuracy rate of the semantic scenes can be increased, and the method is good in generalization performance, provides good assistance to the video scene recognition, and increases the accuracy rate of the video scene recognition.

Description

technical field [0001] The invention relates to the fields of pattern recognition and multimedia information processing, in particular to an audio scene recognition method based on acoustic events. Background technique [0002] At present, with the rapid development of the information society, multimedia information data has shown explosive growth. How to effectively use these multimedia data to serve people's daily life has become an urgent problem to be solved. Multimedia data includes images, audio and other forms. At present, the research and utilization of images has been very extensive, but the research on audio started relatively late, and there are still many technical problems to be solved urgently. [0003] A continuous audio stream usually contains a series of acoustic events, such as speech, laughter, music, etc., and an audio scene refers to an audio segment composed of several temporally adjacent and semantically related acoustic events. . Compared with acou...

AI Technical Summary

Problems solved by technology

Audio scene recognition has such a broad application prospect and urgent market demand, but at present, there are few researches on audio scene recognition at home and abroad, which cannot meet people's needs.

Images