Content-based Audio Retrieval

The rapid development of modern information technology, as well as the rapidgrowth of the amount of information, have enhanced people’s expectations forinformation retrieval. The traditional retrieval methods are all based on key words oftext pattern which are quite incomplete and inefficient. As the audio processingtechnology develops, content-based audio retrieval method has gained more and moreattentions. It can greatly improve the retrieval efficiency of audio signal and avoid themany disadvantages of text retrieval method.The basis of audio retrieval is various audio formats and audio processingtechnology. As for audio formats, the common audio formats are introduced, such asMIDI, PCM and WAV, and their similarities and differences are presented, too. As foraudio processing technology, firstly two kind of audio retrieval models are proposed:Vector Space Model and Probability Model, both of which have certain applicationsaccording to their own features. The usual procedure of audio retrieval is firstlysplitting the audio file into segmentations and extracting corresponding characteristics,and then starting matching and comparing. So far there are mainly two methods:Audio Retrieval Based On Direct Matching and Audio Retrieval Based On HiddenMarkov Model. Then the applications of audio retrieval in music retrieval and audioexample retrieval are analyzed. Lastly, the practical application in AD retrieval isshown to verify the feasibility and superiority of the technology of Content-BasedAudio Retrieval.The innovation of this article is the proposal of AD retrieval based the retrieval ofAD heads. This method can solve the problem of AD retrieval by the real time audioretrieval technology.