Chinese style pop music generation based on recurrent neural network

The existing research on automatic music generation based on recurrent neural network mainly focuses on western classical music, and lacks systematic research on Asian music, especially the Chinese style pop music, which is popular in China at present. According to the rhythm and harmony of Chinese style pop music, this paper designs a music generation system of Chinese style pop music by representing the musical information of Chinese style pop music and constructing a recurrent neural network to learn the specific style from the represented dataset. Firstly, the method uses the polyphony piano transcription technology to convert mp3 format music to midi format music, and carries out the corresponding representation and preprocessing to make the dataset. Then, recurrent neural network model is constructed to model the music style memory and format of the corresponding music data and finally generate music. The results show that the recurrent neural network in this paper can generate Chinese style pop music with the corresponding rhythm and harmony.