Music Regeneration with RNN Architecture Using LSTM

Music generation and restoration are important tasks for audio engineers. It needs to restore the original music that is not fully available currently due to failure or any other reason. The proposed method can be useful for the restoration of any distorted signal. In this paper, the authors have tried to generate and restore classical music, i.e., horn pipe music Initially, the dictionary is prepared for this specific music from which the model extracts the features for learning. The training pattern is developed as an automated one. The model is chosen with a deep learning-based recursive neural network (RNN), where the features are extracted, matched, and reconstructed for this missing or distorted music. The RNN consists of 1024 LSTM modules.