Joint detection method of music and music relative loudness in audio

Abstract

The invention discloses a method for joint detection of music in audio and the relative loudness of music, which comprises acquiring the audio to be detected and preprocessing it to obtain audio features; inputting the audio features as a network input sequence into a dual-task music detection network model, Output three prediction scores for each time step; based on the prediction score ≥ the set threshold or < the set threshold is marked as 1 or 0, the prediction score of each time step is converted into a size of (3,) 0‑1 Prediction vector, using the prediction vectors of all time steps to form a music relative loudness detection sequence; combining the second and third positions of the prediction vector with logical or operation to obtain a 0-1 prediction vector with a size of (2,), and using logical or operation to obtain all The prediction vector forms a music detection sequence; the music relative loudness detection sequence and the music detection sequence are converted into readable annotations, and an audio annotation file including start time, end time and category is obtained.

Description

technical field [0001] The invention relates to music detection technology, in particular to a joint detection method of music in audio and the relative loudness of music. Background technique [0002] Music detection is a kind of music information retrieval task. Music detection refers to given an audio file to determine whether it contains music fragments and the start and end positions of music fragments. Two main applications of the music detection task are: (1) automatic indexing and retrieval of auditory information based on audio content; and (2) monitoring of music for copyright management. In addition, music detection can also serve as an intermediate step for other music information retrieval tasks and improve the performance of other music information retrieval algorithms. [0003] The music detection task initially only focuses on distinguishing the sound information of the two categories of music and human speech in audio files (the other sounds are marked as n...

AI Technical Summary

Problems solved by technology

[0004] Most of the existing music detection technologies still focus on the distinction between human voice and music sound, without considering the estimation of the relative loudness of music, which makes the effect of music detection not good

Images