Music generation and human voice conversion based on LSTM

Music is closely related to human life and an important way for people express their feelings in life. Deep neural networks have played a significant role the field of music processing. There are many different network models implement deep learning audio For general networks, there problems such as complex operation slow computing speed. In this paper, we introduce Long Short-Term Memory (LSTM), which circulating network, realize end-to-end training. The structure simple can generate better sequences after training model. After generation, voice conversion understanding inserting lyrics pure music. We propose segmentation technology segmenting fixed length voice. Different notes classified through piano without considering scale correlated with voices get. Finally, transformation, generated output Experimental results demonstrate that proposed scheme successfully obtain from by LSTM.