Research on Automatic Piano Music Transcription Algorithm Based on CNN

With the development of deep learning technology, automatic music transcription has gradually become an important research direction in the field of music information processing. In particular, manual transcription is no longer sufficient to meet the large-scale automation needs of complex piano music. This paper proposes an automatic piano music transcription model that combines Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Multilayer Perceptron (MLP). The model extracts spatial features from spectrograms using CNN, processes the temporal dependencies of note sequences with RNN, and optimizes the relationships between notes with MLP to achieve accurate note recognition and coherent score generation. Experiments were conducted on multiple piano music datasets, and the results show that the improved model outperforms traditional methods in terms of note recognition accuracy, recall, and F1 score, with an average F1 score exceeding 90%. This model provides an efficient solution for music transcription tasks and demonstrates the potential of deep learning in complex music information processing.