An Improved Music Composing Technique Based on Neural Network Model

Traditionally, music was considered an analog signal that had to be made by hand. In recent decades, has been highlighted technology can autonomously compose a suite of without any human interaction. To achieve this goal, article suggests autonomous composition technique based on long short-term memory recurrent neural networks. Firstly, the collection is split into sequences unit time, and Meier cepstrum coefficients audio are retrieved as features during preprocessing. Secondly, training samples composed feature vectors processed data were trained predicted short- long-term models. Finally, generated sequence spliced fused get new music. This designs performs experiments demonstrate our results promising. From experimental results, work gained model maximum accuracy 99% lowest loss rate 0.03.