A Research On Audio Sample Retrieval Technology

Into the 21st century, communication technology, Internet technology and multimedia technology has been rapid development of information and data in multimedia data has greatly exceeded the amount of text data and still maintained rapid growth. Audio, video and image is the main part of multimedia information, in order to make full use of audio information, people started to pay attention audio information retrieval technology. In relation to images and video and audio data have the characteristics of unstructured, a feature that allows audio information retrieval over images and video information retrieval more difficult. The audio information from the existence of formal terms, there are two, one is being stored in some medium, such as CD-ROM or tape, the other is real-time playback, such as broadcasting. This has resulted in the audio retrieval of the sub-offline and online. Another Audio Retrieval can also be divided into that level and semantic level. Therefore, to retrieve different forms require different audio retrieval method. Audio Information Retrieval technology development time is not very long, at present there are still many problems to be resolved. As a whole, practical retrieval systems are also very small, audio retrieval technology is still in its infancy stage of research.In this paper, research shows that level audio search, retrieval for audio samples and conduct the following research work:1, for noise-sensitive audio sample retrieval problem, a threshold-based adaptive histogram of the audio retrieval method. In the audio feature vector quantization, according to the characteristics of the audio features optimize the method for vector quantization codebook. Dynamic histogram analysis of the shortcomings in noisy situations, combined with a search feature coding matching threshold adaptive control. Experimental results show that the method has better noise robustness.2, for the deterioration of the audio retrieval robustness issue, the division of sub-template retrieval method. This method will draw the audio is divided into several sub-templates, and use the sub-templates registered to slide the window to control whether or retrieval. Combination of experimental analysis of the target audio defects occur in different parts of the algorithm and finally compare the overall retrieval methods show that the method of incomplete robustness.