Human Emotion Recognition System using Residual Connections

The goal is to use a Multi-Branch Cross-Connected Convolutional neural network to construct a new model for expression detection based on a novel face decomposition. CNN, residual connections, and Multi Networks are all included in this strategy. Other approaches may not be able to extract as many features as this method. This method adds a shortcut link, resulting in a highly smooth data or feature flow, as well as good feature extraction. Some methods are not that great in extracting complete features; this will be sorted out in this method. This method can recognize all the emotions, the emotion which shows the highest value is our output. The data flow in Single-structure CNN is not smooth, and feature extraction is insufficient. The proposed method can perform well in terms of emotion recognition and has a high level of accuracy. More fluid data transfer between networks improves feature extraction and increases emotion recognition performance.