Hand Gesture Recognition Based on Flex Sensors with Enhanced Focal Loss

A novel hand gesture detection algorithm for a smart Human Computer Interface (HCI) is developed in this study. Flex sensors are deployed in a smart glove for the collection sensory data for finger movements. Based on the sensory data, a modified centernet algorithm is presented for gesture detection and classification. To address class imbalance and enhance sequence-level precision, a Focal Loss function is incorporated into the model training. The algorithm has the advantages of accurate gesture detection and classification with low computational complexities. The algorithm can be deployed in an embedded system for effective online inference. Experimental results reveal that the proposed algorithm is an effective alternative for applications based on smart gloves requiring fast and accurate gesture detection recognition with low deployment overhead.