While AI has been known to recognise musical melodies, scientists have further improved the technology to specifically identify musical notation.
In an aim to enhance online music education, Ting Zhang from the Academy of Arts at Shangluo University, Shaanxi, China, developed a new AI tool that recognises musical notations.
Classical music lovers particularly often face challenges with digital music instruction due to the platform’s limitations of not being able to aptly detect and interpret musical notation.
This prompted the development of the new musical AI tool which leveraged machine learning (ML) and image processing techniques to help online users understand musical concepts more accurately.
The accuracy rate of the new AI model for music notation identification is estimated at 97%.
The study explained that it investigated how to improve the pulse-coupled neural network (PCNN) in image processing for spectral image segmentation.
PCNN is an artificial neural network inspired by the functions of biological neurons that respond to particular stimuli.
Unlike traditional online music learning, the PCNN significantly enhanced the digital segmentation of musical symbols within an image of a musical score according to TechXplore.
Image Segmentation is 98% Accurate
Conventional digital music learning relies on simplified digital representations of musical notation without specific guidance or additional explanations.
The music notation-based AI tool integrated oblique spectral correction into the system which dissects the image into parts and precisely spots the difference between musical symbols.
According to the research, the results showed that the image segmentation method had the highest accuracy, a staggering 98%.
The accuracy level of spectral tilt correction was claimed to be as high as 98.4%, presenting successful pre-processing results.
“When combined with the improved convolutional neural network (CNN), the average accuracy of music theory symbol recognition was about 97% and the highest score of music majors was improved by 16 points,” said Zhang.
“The method can effectively improve the teaching effect of online training in music education and has certain practical value.”
How Does the Music Notation AI Tool Work?
The new AI tool optimises image processing and machine learning (ML) techniques to interpret musical notations accurately.
Zhang deploys a two-scale descent method to achieve oblique spectral correction. According to Science Direct, oblique projection is a chemometric technique which removes multi-scattering effects in close-range hyperspectral images of vegetation scenes.
Essentially, the AI tool works by removing the non-useful part of the image while reducing the loss of information which helps in precision detection.
The tool combines PCNN to separate musical symbols from images of scores, even when distorted or tilted. An optimised CNN then recognizes the segmented symbols with high accuracy.
This real-time feedback system improves learners' understanding of pitch, rhythm, and foundational music theory concepts, bridging the gap between online and traditional music instruction.
What is an Image Processing Technique?
Image processing is a technique that improves or analyses digital images using computer algorithms to yield information or improve their quality. These images can be captured from multiple sources, such as cameras, satellites, or sensors.
Image processing tools can be used to analyse the valuable information prompted by a user or query. These techniques can filter, reduce noise, detect edges, or manipulate colours to extract meaningful insights from the images.
According to Science Direct, image processing is one of the leading ICT technologies used in many applications such as crop monitoring and yield estimation in Smart Farming.
