Central AsianResearch Index

Development of a web application for intelligent analysis and comparison of musical compositions using audio feature extraction and neural networks

The article presents the development results of the «tunefinder» web application designed for the intelligent analysis of musical compositions. The system enables users to analyze audio tracks by extract-ing quantitative musical features such as tempo (BPM), key, and rhythmic structure using advanced algo-rithms and neural networks. The solution integrates convolutional-recurrent neural networks (CRNNs), the Krumhansl-Schmuckler algorithm with the Aarden-Essen profile, and the RhythmExtractor2013 from the Essentia library. The architecture is built on a modular and scalable design with a Gradio-based in-terface and RESTful API for efficient data exchange. MongoDB is used for flexible NoSQL data storage. The application supports external integration with Shazam and Spotify to enhance metadata analysis. Op-timization methods such as asynchronous processing, result caching, and server-sent events (SSE) ensure high performance. The service serves musicians, educators, and analysts, providing tools for single-track evaluation, comparative key analysis, and interactive visualizations.

Сведения

Опубликовано
2025 · Т. 98, № 4, С. 110-122
Язык
Английский
ISSN
2709-4693 · 2709-4707
Издатель
National Engineering Academy of the Republic of Kazakhstan