Music generation method based on facial expression recognition and recurrent neural network

Abstract

The invention discloses a music generation method based on expression recognition and a recurrent neural network. The method comprises the following steps: 1) obtaining music audio data and characterexpression data; 2) classifying and marking the data; 3) processing audio data and image data; 4) initializing an RNN-RBM neural network; 5) training the RNN-RBM neural network; and 6) identifying thefacial expression by using VGG19 + dropout + 10crop + softmax. And 7) inputting the recognized emotion information into the trained RNN-RBM network to obtain final music generation The emotion regulation method combines facial emotion recognition and AI music generation, can generate music according to the emotion of the person, achieves the purpose of emotion regulation, and has high practical application value.

Description

technical field [0001] The invention relates to the field of computer technology and the field of digital music generation, in particular to a music generation method based on expression emotion recognition and cyclic neural network. Background technique [0002] Music has a subtle influence on people's body and mind. With the development and progress of the Internet and cloud music, music takes up more and more time in people's daily life, and silently regulates people's physical and mental health. We can deeply feel the role of music in our daily life. For example, programmers will improve their programming efficiency when they listen to music, bodybuilders are used to using music to adjust their fitness rhythm, drivers use music to improve their concentration while driving, and so on. Listening to the right music on a suitable occasion can also greatly stretch people's body and mind. For example, listening to a passionate symphony can release people's low mood when they ...

AI Technical Summary

Problems solved by technology

[0004] In order to make up for the defect that existing music cannot meet the individual needs of listeners on specific occasions, the present invention combines facial expression emotion recognition and cyclic neural network music generation technologies, and proposes a music generation based on facial expression emotion recognition and cyclic neural network method, wherein facial expression emotion recognition adopts Visual Geometry Group-19 convolutional neural network model (VGG19 for short), and the music generation method of recurrent neural network adopts recurrent neural network-restricted Boltzmann machine (Recurrent Neural Network-Restricted Boltzmann Machine) , referred to as RNN-RBM) algorithm

Images