Method for automatically detecting similarity level of audio signals and system of method

Abstract

The invention discloses a method for automatically detecting the similarity level of audio signals and a system of the method. The method includes the steps that the two segments of audio signals to be detected are obtained; frequency-domain analysis is performed on the audio signals, and all frequency-domain energy local peak value positions are found; frequency-domain energy local peak values are connected to form a pitch trail; according to the frequency-domain energy local peak value positions, a significant value of each audio frame is calculated, and an average significant value is obtained; according to the average significant value, a threshold value of phonic part judgment on the pitch trail is calculated; the part, exceeding the threshold value, of the average significant value is removed to obtain main melody sequences of the two segments of audio signals; a similarity matrix of two main melodies is obtained; a dynamic planning local warping algorithm is applied to binarization processing, so that a binary matrix is obtained; then the similarity level of the two segments of input audio signals is judged. By the adoption of the method or the system, detection efficiency is improved for the similarity level of the audio signals, the problem of plagiarism in audio spread is solved, and accurate and comprehensive protection is provided for right management of digital audios.

Description

technical field [0001] The invention relates to the field of music plagiarism detection, in particular to an automatic detection method and system for audio signal similarity. Background technique [0002] With the rapid development of the Internet and software development technology, the cost of music dissemination and the threshold of music writing have become lower and lower, and audio similarity comparison has become an important research topic. In the research of music information retrieval, music similarity calculation is one of the key topics, which studies how to calculate the similarity in the content of two different audio pieces. The definition of "similarity" between music usually contains a lot of subjective elements, but there are also some quantifiable entry points, such as the degree of similarity in the direction of melody. There are two opinions in the judgment criteria of music plagiarism , One is that the similarity of more than 8 bars is regarded as pla...

AI Technical Summary

Problems solved by technology

[0004] However, the main problem of the prior art is that the effect of judging music plagiarism is not comprehensive

More music plagiarism only needs to be calculated in the local part of the music. The dynamic time warping algorithm calculates the similar distance from (0, 0) to (Am, Bm) and cannot well represent the relationship between two pieces of music. Similarity difference, so in the disordered locally similar music, this method does not show significant similarity difference well. If you want to use local interception for processing, it will undoubtedly increase the time overhead. In the quick judgment Under the technical requirements of plagiarism, it cannot be the first choice

At the same time, there are various definitions of sequence distance. This method based on linear comparison has no obvious gradient in the calculation of similarity. In other words, the fitting of similarity should not be a simple growth curve. The definition of distance must be affect the judgment

Images