• Sanskar Tidke
• Shrey Shah
• Rudrang Wadal
• Shruti Narwane

This paper explores the application of Long Short- Term Memory (LSTM) networks, a type of Recurrent Neural Network (RNN), for automatic music generation. We address the technical challenges involved in this task and propose an algorithm for generating musical sequences. In this research, we use MIDI files to represent the music data. MIDI stands for Musical Instrument Digital Interface, and it’s a common format for storing musical information. This format allows our system to process and analyze the music efficiently. We’ll explain exactly how we prepare the music in the MIDI files for training the LSTM model. This involves steps like reading the data, processing it, and converting the musical sequences into a format the model can understand. The core of our system is a special kind of neural network called a Long Short-Term Memory network, or LSTM for short. We’ve designed this network with a single layer specifically to capture the long-term patterns in complex music with multiple melodies playing at once (polyphonic music). We’ll break down the structure of this LSTM network layer by layer, explaining how the different parts are connected to create the overall architecture. To provide insights into the training process, we analyze the distribution of weights and biases across the network layers. Additionally, we present a detailed evaluation of the model’s performance, including loss and accuracy metrics measured at various training stages and batches. This research contributes to the field of music information retrieval (MIR) by exploring the potential of LSTMs for automatic music generation.

Music, Melodies, RNN, LSTM, Neural Net- work, MIDI.

"Sound Generation using Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 4, page no.k475-k481, April-2024, Available :http://www.jetir.org/papers/JETIR2404A64.pdf

"Sound Generation using Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 4, page no. ppk475-k481, April-2024, Available at : http://www.jetir.org/papers/JETIR2404A64.pdf