Institutional Repository of Vilnius University
Muzikos platformos su žanrų atpažinimo funkcionalumu kūrimas /
Abstract
dc:descriptionThe motivation for this work arose from the need to automate the classification of rapidly growing digital music libraries to improve user experience on platforms such as Spotify or SoundCloud. The aim of the work is to investigate music audio signal classification methods and to create a prototype music platform with automated genre recognition functionality. To achieve this goal, the study employs the mel spectrogram method, which transforms a one-dimensional audio signal into a two-dimensional graphical form that best matches human biological frequency perception. A convolutional neural network (CNN) architecture was implemented for classification, consisting of three convolutional blocks capable of hierarchically extracting features from the texture of the spectrograms. To ensure model stability and avoid overfitting, the initial GTZAN dataset was expanded to 18,971 samples by applying a 3-second segmentation strategy with 50% overlap. Experimental results showed that the developed system achieved a high theoretical accuracy of 93.12% on the test data set. The resulting classification model was integrated into a web application developed using Laravel, Vue.js, and Python technologies. This platform provides users not only with an automatic genre identification function with detailed probability visualization but also with practical tools: an integrated player, a playlist system, and the ability to change playback speed. Although practical testing in a real-world environment revealed that the system's accuracy depends on the specifics of the analyzed audio fragment, the trained model confirmed its effectiveness in recognizing complex musical features.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mikalauskaitė, Austėja,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277690024&prefLang=en_US