A method and device for automatic notation based on sound source separation

By pre-training multiple sound source separation models and combining automatic notation technology, the problem of the difficulty of accurately separating the sound of instruments in a multi-sound source environment is solved, and multi-instrument recognition and separation are realized without the need for specific training data.

CN117995140BActive Publication Date: 2025-03-18BEIJING UNIV OF CIVIL ENG & ARCHITECTURE
2 Cites 2 Cited by

Patent Information

Application Number
CN202311858130.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-29
Publication Date
2025-03-18
Estimated Expiration
2043-12-29

AI Technical Summary

Technical Problem

Traditional automatic notation technology is difficult to accurately separate the sounds of different instruments in multiple sound source environments, and it is difficult to obtain training data, resulting in the inability to effectively train the sound source separation model.

Method used

By pre-training multiple sound source separation models, each model corresponds to one instrument, using the target audio file input by the user to separate the sound source, output the audio sequence of the target instrument or all the instruments, and perform noise suppression and optimization, and finally generate the music score through automatic notation technology.

Benefits of technology

It realizes accurate identification and separation of target instruments in a multi-instrument environment, improves the effect of audio separation and recognition, and trains the target sound source separation model without specific audio samples, solving the problem of difficulty in obtaining training data.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present application provides an automatic notation method and device based on sound source separation. The automatic notation method based on sound source separation provided by the present application includes: according to the target audio file input by the user, using the target sound source separation model among multiple sound source separation models, outputting the audio sequence of the target instrument or all instruments contained in the target audio file, and the target sound source separation model is one or more; performing noise suppression on the audio sequence and optimizing the audio sequence; according to the optimized audio sequence, outputting the corresponding music score through the automatic notation technology, obtaining the data set of the target instrument by pre-processing the data set, and then using the data set of the target instrument to train the target sound source separation model, solving the problem of difficulty in obtaining training data and the inability to train the sound source separation model; outputting the music score of the target audio file through the target sound source separation model and the automatic notation technology, achieving the effect of multi-instrument recognition and separation.
Need to check novelty before this filing date? Find Prior Art