Figure 1.

This is a summary of the gait recognition model we developed. (a). The raw video is collected from an RGB-D sensor. The skeletal joint rotations are provided by the sensor’s technology, which we preprocess and define in the ST-Graph. (b). The deep learning model is constructed according to the arrangement ST-Graph. (c). For each layer, the model outputs the feature maps of size (T, 128), (T, 256), and (T, 512). (d). At inference, the model makes the class prediction according to the Maximum Likelihood Estimation (MLE) of the model after training.