GenreNet: A Deep Based Approach for Music Genre Classification

Neural network-based Music Genre Classification is a key component in helping users narrow down the selection of songs and listen to music in a certain genre. Audio segmentation has been done as a preprocessing step on the available audio files using the benchmark dataset “GTZAN”. To improve accurac...

Full description

Saved in:
Bibliographic Details
Published in:SN computer science 2024-12, Vol.5 (8), p.1135
Main Authors: Bala Ganesh, N., Bhuvaneswari, M. S., Bhagavathi Sankar, K., Ganesh, P.
Format: Article
Language:English
Subjects:
Accuracy
Acoustics
Audio data
Classification
Computer Imaging
Computer Science
Computer Systems Organization and Communication Networks
Data Structures and Information Theory
Datasets
Deep learning
Error analysis
Fourier transforms
Frequencies
Genre
Information Systems and Communication Service
Machine learning
Musical performances
Neural networks
Optimization
Optimization techniques
Original Research
Pattern Recognition and Graphics
Segmentation
Signal quality
Software Engineering/Programming and Operating Systems
Spectrograms
Support vector machines
Vision
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Neural network-based Music Genre Classification is a key component in helping users narrow down the selection of songs and listen to music in a certain genre. Audio segmentation has been done as a preprocessing step on the available audio files using the benchmark dataset “GTZAN”. To improve accuracy, audio segmentation divides a 30-s audio recording into ten 3-s audio files. Segments are utilized to create spectrograms, which are then used to extract features. A spectrogram is a visual depiction of a signal’s frequency band. Applying techniques like magphase and Short Time Fourier Transform to the audio data enhances the quality of the spectrograms. Adaptive Moment estimation, Nesterov-accelerated Adaptive Moment estimation (NADAM), Adaptive Moment estimation with Maximum, and Root Mean Squared Propagation are among the optimization techniques used for feature weight optimization to increase the classification rate. NADAM is found to be effective based on performance measures. The proposed GenreNet model is compared with the state-of-art techniques and performance measures such as Accuracy, Precision, Recall, Root Mean Square Error, F1 Score, Loss and Sensitivity have yielded positive results.
ISSN:2662-995X
2661-8907
DOI:10.1007/s42979-024-03493-x