A real-time audio comparison method based on Walsh-Hatta code transform

Abstract

The invention belongs to the technical field of digital audio processing, specifically an audio real-time comparison method based on Walsh-Hadamard transform. The method comprises the steps: S1, feature extraction: carrying out the preprocessing, audio frame reading, Walsh-Hadamard transform, audio frame characteristic calculation and audio characteristic generation of two to-be-compared audio signals, wherein the characteristic vectors fi of all audio frames are arranged in a matrix F = [f1, f2, f3,..., fM]; S2, feature matching: setting the characteristic matrixes of the two to-be-compared audio signals as F and conjugate F, comparing the corresponding characteristic vectors of the characteristic matrixes F and conjugate F, and enabling the distances of all frames of the two to-be-compared audio signals to be averaged, wherein the mean value serves as the distance between the two to-be-compared audio signals; S3, similarity determination. The method is higher in comparison fault tolerance, and can achieve accurate comparison under a condition that the audio signals are affected by noise or a condition of sound processing. Meanwhile, the method is simple in algorithm, is high in processing speed, and can be widely used for the program monitoring, advertisement monitoring and management of a broadcasting station.

Description

technical field [0001] The invention belongs to the technical field of digital audio processing, and relates to a method for comparing two audio signals, specifically a real-time audio comparison method based on Walsh-Hatta code transformation, which can be applied to real-time comparison of broadcast audio right. Background technique [0002] Today, audio broadcasting has become the most popular medium of publicity and entertainment. The program broadcasting of radio stations and the management of media resources all involve the need for audio comparison, that is, to compare two audios to determine whether they are the same or similar. For example, in program monitoring, it is necessary to receive and query the audio signals actually transmitted and broadcasted online in real time, which involves comparing the received audio stream with the original audio to be broadcast, so as to judge the entire open-loop broadcast channel Is it normal? Another example is in advertising...

AI Technical Summary

Problems solved by technology

In this case, the traditional comparison method is often not applicable, because the broadcast audio comparison needs to determine whether the broadcast audio content itself is the same, rather than whether the auditory effect is consistent, that is, even if a certain audio is disturbed by certain noise, spectrum After adjustment and other processing, the algorithm should still determine that the audio is similar to the original audio (or similar to each other)

Although comparison methods that are not sensitive to audio changes have also appeared in the literature, such as the comparison method based on Philips audio fingerprints, in these methods, the distance between two audio before and after processing is not much different from the distance between two audio that are really completely different. Therefore, it is easy to be confused in discrimination, especially when there is strong noise interference or a large change in audio processing, discrimination errors will occur; on the other hand, in terms of computational efficiency of comparison, the computational complexity of existing methods is relatively low. High, the time required for comparison is relatively long, which is not conducive to real-time processing

Images