Classification of Assamese regional songs using a hybrid model of machine learning and deep learning

Purpose The purpose of the study on Regional music serves as a vital medium for preserving cultural identity, traditions and collective memory. In Assam, a state in Northeast India, musical forms such as Bihu, Borgeet, Goalpariya, Kamrupi and Naam represent the region’s diverse cultural and spiritual heritage. Despite their significance, systematic computational studies on Assamese regional music remain scarce. Design/methodology/approach This paper presents an approach for automatic classification of Assamese regional music using a hybrid deep learning and machine learning technique. Audio recordings were transformed into spectrograms to capture both frequency and temporal features of the signals. Findings A hybrid CNN + KNN model was developed, which yielded an accuracy of 0.71, with precision of 0.76, recall of 0.71 and F1-score of 0.70. The hybrid model demonstrated a balanced performance across evaluation metrics, indicating improved generalizability. Originality/value To the best of the authors’ knowledge, this study contributes to music information retrieval (MIR) and digital heritage by being one of the first systematic attempts to classify Assamese regional music through artificial intelligence-driven methods. The proposed framework can support digital archiving, recommendation systems and cultural preservation, ensuring that Assamese traditions are safeguarded in the digital era.