Superposed CNN-RBN joint network-based gesture recognition system

The hand is a non-rigid item with a wide range of motions, making gesture identification more complex. The organization and recognition of a frame still images lie at the heart of dynamic gesture recognition. As a result, this paper focuses mostly on static gesture identification. There are currently various issues with gesture recognition, including as accuracy, real-time capability, and resilience. In order to access address aforementioned issues, this research proposes a gesture recognition network that combines CNN and RBM. It primarily employs a superposed network of numerous RBMs for unsupervised feature extraction, which is then merged with CNN supervised feature extraction. These characteristics are then blend to collocate them. This modelling findings suggest that the advance superposed network performs superior in detecting simple and complex backdrop gesture samples, that gesture sample detection in complex backgrounds still needs to be improved.