Semisupervised Deep Features of Time-Frequency Maps for Multimodal Emotion Recognition

Figure 3

The procedure used to optimize the parameters of SSDR and classifier.