| abstract |
The embodiment of the invention discloses a music generation method and device based on a neural network, the method comprises: acquiring target music data, performing clustering processing on the target music data to obtain a global music sequence; acquiring the first music sequence, and acquiring the global music sequence Entering the length and time of the first music sequence to memorize the LSTM network model, the first network layer based on the LSTM network model confirms the sequence context information corresponding to the first music sequence and outputs the target note corresponding to the sequence context information; and inputs the target note into the LSTM network. The second network layer of the model obtains input duration information of the target note through the second network layer; generates a second music sequence according to the input duration duration information and the target note of the target note, and obtains the synthesized music based on the second music sequence. By adopting the embodiment of the invention, the efficiency of music synthesis can be effectively improved, the interest of synthetic music can be enhanced, and the applicability is stronger. |