Music reorganization-based music score automatic formation method

Abstract

The invention discloses a music reorganization-based music score automatic formation method. The method includes the following steps that: (1) audios are recognized, the changes of the whole frequency spectrum of the audios are tracked, and sound emitting trends are checked in real time; (2) the frequency spectrum changes of each pitch are tracked, and pitches requiring sound emitting are checked in real time; (3) the frequency spectrum changes of pitches of which the sound emitting has been completed are further tracked, and whether judgment on sound emitting of the previous pitches is wrong is checked; and (4) the speed, tone and note type of a music score are estimated according to sound-emitting pitch data and sound-emitting time data obtained in the above steps, so that the music score can be composed. According to the music reorganization-based music score automatic formation method, an original music score can be estimated reversely according to pitch reorganization results, and thus, audio music can be automatically identified, and a music score can be automatically formed. The music reorganization-based music score automatic formation method has the advantages of simple calculation, robustness in sound emission number, compatibility with requirements of a variety of musical instruments as well as being efficient and reliable. The music reorganization-based music score automatic formation method can be applied to mobile phone software and embedded equipment, and can be also applied to occasions such as composition creation automatic music score formation and musical instrument playing exercise check.

Description

technical field [0001] The invention relates to a method capable of recognizing music and automatically forming scores, and belongs to the technical fields of music pitch recognition, polyphonic recognition, automatic score formation and the like. Background technique [0002] Pitch recognition technology has a wide range of uses and can be used in tuners, music recognition, and audio file conversion. In the tuner of musical instruments, the pitch recognition technology of single tone is mainly used, and the fundamental frequency of the sound is identified by the harmonic peak method to check whether the pitch frequency of the musical instrument is shifted, but the pitch recognition achieved by this technology Only a single tone can be recognized, if several pitches are sounded at the same time, it will not be recognized. In the application of music recognition, the pitch recognition technology will be used to roughly recognize the melody hummed by the user, and use this as...

AI Technical Summary

Problems solved by technology

Among them, the harmonic peak method is aimed at the analysis of frequency-domain harmonics. The advantage is that the method is intuitive and achievable, and the amount of calculation is not large. However, because it relies on the harmonic with the largest energy as the starting analysis point, it cannot analyze the situation of multi-tone recognition. At the same time, due to the The pronunciation of musical instruments is often harmonically shifted, and the analyzed fundamental frequency is often inaccurate

The parallel processing method is for audio analysis in the time domain. According to the principle that the pitch and the harmonics of the pitch are periodically and regularly superimposed on the time axis, the distance between the two peaks is estimated, and the audio period is calculated from this. The advantage is the calculation and implementation are relatively simple, the disadvantage is that the result is not stable

The wavelet analysis method is to use wavelet transform to analyze the frequency domain features in depth and extract the pitch. Its advantage is that the accuracy is high, and the disadvantage is that the amount of calculation is too large.

[0004] The multi-tone estimation technology has always been a difficult point in music recognition. Since Moore, a scholar at Stanford University in the 1970s, began to study it, Osaka University's "Sensory Information Extraction of Music Signals" research project in the 1980s, Hawley of the Massachusetts Institute of Technology and The Martin project, the Kashino project of the University of Tokyo, the Sterian project of the University of Michigan, and the Douglas Nunn project of Dalamo University have all made corresponding research breakthroughs, but there has never been a single operation that is simple, efficient and credible, and has a robustness in the number of pronunciations. The method of being compatible with various musical instruments can be applied to non-laboratory small computing device software, such as smart phones, smart tablets, embedded devices, etc.