A Speaker System Based On CLDNN Music Emotion Recognition Algorithm

Inhalt:
In this paper, We have designed a speaker system that can automatically recognize the emotion of playing music. The core of the entire system is a music emotion recognition algorithm based on convolutional long short term memory deep neural network (CLDNN). In addition, We use the PMEmo emotional music dataset as the training data of the neural network, which contains 794 music clips and their corresponding emotional labels. The system proposed in this article runs on a hardware platform with Raspberry Π 4b as the main controller. After we say the wake-up word "SNOW" to the speaker system, the system will automatically recognize the name of the song and play it. While playing the song, the mood of playing the music will be automatically recognized by CLDNN algorithm. According to different emotion categories, different lighting effects will be presented in the system lighting module composed of WS2812 lamp beads. The music emotion classification model that forms the core of the system is a deep neural network composed of CNN+LSTM+DNN. The input of the neural network is the MFCCs characteristics of music, and the output is the music emotion coordinates under the V-A coordinates. The recognition accuracy of the whole model is about 90%, which achieves an ideal effect.