Similarity measurement method for audio-frequency fragments

Abstract

The invention discloses a method of measuring similarity degree among the acoustical frequency fragment. The present technique represents the whole acoustical frequency fragment with the acoustical frequency characteristic owing to not considering the difference of the specific content in the acoustical frequency fragment, so the present technique doesn' t measure the similarity degree of the acoustical frequency content effectively. Pointing to the above problem, the invention divides the acoustical frequency fragment in to two layers: the acoustical frequency unit and the acoustical frequency fragment. In the acoustical frequency unit procedure, the acoustical frequency unit is defined series of acoustical frequency frames with similar acoustic fidelity. Firstly, the acoustical frequency fragment is divided into series of the acoustical frequency units and the similar degree of the acoustical frequency unit is measured; what' s more, in the acoustical frequency fragment procedure, the similar degree of two acoustical frequency fragments is measured and modeled a weighted dimidiate graph based on the measuring result of the acoustical frequency unit; at last, the similar degree of two acoustical frequency fragments is measured with optimum matching. The invention provides higher search accuracy, which shows the important effect of the acoustical search technique in the information retrieval.

Description

technical field [0001] The invention belongs to the technical field of audio retrieval, and in particular relates to a method for measuring the similarity between audio clips. Background technique [0002] With the continuous increase of multimedia materials and applications, audio analysis and retrieval technology is becoming more and more important, and audio clip retrieval is an important form of the above technology, that is, how to automatically retrieve an audio clip from the audio library given by the user Find similar audio clips and sort them in descending order of similarity. Existing audio retrieval technologies generally extract audio features from audio clips, and then use these features to measure similarity, and perform retrieval based on the measurement results. Because this method does not consider the difference in the specific content of the audio clips, but uses audio features to represent the entire audio clip, it cannot effectively measure the similari...

AI Technical Summary

Problems solved by technology

This method is based on the statistical characteristics of the entire audio clip, so it cannot describe the content change characteristics in the audio clip, thus limiting the accuracy of audio retrieval

Images