Content-Based Hybrid Deep Learning Model for Classical Music Recommendation

Indian Classical Music (ICM) is characterized by intricate melodic and rhythmic structures, making its analysis and recommendation a challenging task. Unlike Western music, ICM relies on raga based improvisation, requiring specialized feature extraction techniques for meaningful representation. This study proposes a hybrid deep learning framework combining Convolutional Autoencoders (CAE) and Long Short-Term Memory (LSTM) networks for content-based recommendation ICM. The CAE extracts spectral features from Mel-Spectrograms, while the LSTM learns temporal dependencies to enhance feature representations. The learned embeddings are utilized for music classification (classical vs. non-classical) and recommendation based on cosine similarity of extracted features. Performance is measured using Mean Precision@K and Mean Reciprocal Rank demonstrating the effectiveness of the proposed model in retrieving similar classical compositions.