Real-Time Human-Music Emotional Interaction Based on Deep Learning and Multimodal Sentiment Analysis

Figure 8

Samples of JCMD Dataset (Jiang’s classical music dataset). 250 normalized clips of polyphonic classical music are collected and labelled. Three discrete labels -1, 0, and 1 are used to label different dimensions in the PAD 3D emotion model, which is then used to train and verify the model.