Research on Music Style Classification and Recognition Algorithm Based on Artificial Intelligence

Traditional music style classification methods rely heavily on the subjective judgment of professional musicians and are difficult to handle large-scale music datasets. With the rapid development of digital music and Artificial Intelligence (AI) technology, this study utilizes deep learning techniqu...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2025 International Conference on Digital Analysis and Processing, Intelligent Computation (DAPIC) s. 272 - 276
Hlavní autoři: Yan, Lin, Huang, Xiwen, Long, Yin
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 26.02.2025
Témata:
Accuracy
Artificial intelligence
Classification
Classification algorithms
Convolutional neural networks
Data mining
Feature extraction
Long short term memory
Long Short Term Memory Networks
Music
Music Style
Recognition
Support vector machines
Timbre
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Traditional music style classification methods rely heavily on the subjective judgment of professional musicians and are difficult to handle large-scale music datasets. With the rapid development of digital music and Artificial Intelligence (AI) technology, this study utilizes deep learning techniques, especially a hybrid model of Convolutional Neural Networks (CNN) and Long Short Term Memory Networks (LSTM) (CNN-LSTM), to achieve automatic classification and recognition of music styles. Firstly, the characteristics of music are comprehensively analyzed, and the key features such as rhythm, melody, harmony and timbre are selected as the classification basis. Rhythm features are extracted by calculating beat intensity and beat speed; Melody feature uses pitch tracking algorithm to obtain pitch distribution and melody outline; Harmony features are extracted by analyzing chord types and harmony complexity; Timbre features are quantified by spectral centroid and spectral flux. In algorithm design, CNN-LSTM hybrid model combines the advantages of CNN in local feature extraction and LSTM's specialty in dealing with long-term dependence of sequence data. CNN part extracts the spectral features of audio layer by layer through convolution layer and pooling layer, while LSTM part captures the time sequence information in these feature sequences. In addition, the attention mechanism is introduced, so that the model can automatically pay attention to the key features most related to music style, and further improve the classification accuracy. The experimental results show that the algorithm has achieved remarkable results on music data sets including rock, jazz, classical, pop and other styles. Compared with traditional algorithms such as SVM and random forest, CNN-LSTM hybrid model shows superiority in accuracy, recall and F1 score, reaching 92.5% accuracy, 91.8% recall and 92.1 % F1 score. This study not only provides a new perspective and method for music style classification and recognition, but also provides strong support for music research, recommendation system and copyright management.
DOI:10.1109/DAPIC66097.2025.00056