Automatic American sign language prediction for static and dynamic gestures using KFM-CNN

For human–computer interaction, one of the most important tools is Sign Language Recognition in which one of the significant research topics is static Hand Gesture (HG) and dynamic Hand Gesture Recognition (HGR) of American Sign Language (ASL). However, recognizing them was challenging owing to improper contour models. Thus, this paper proposes a new model for ASL Static and dynamic HGR utilizing Deep Learning (DL). Primarily, the videos from the dataset are collected and transformed into image frames and then given to noise filtering utilizing an Alternative Window Size-Lone Diagonal Sorting Algorithm. The RGB noises are removed in noise filtering and given to Canopy Algorithm-based Minibatch K-Means Clustering clustering of static and dynamic gesture images. After that, by utilizing the YCbCr Palm point and Finger speed-based Threshold in the Region seed Grow Segmentation algorithm, the clustered images are segmented that segments both palm and fingers. Then, the features are extracted, and these features are given to the Kohnen Feature Mapping-based CNN classifier. From the Classifier, HG-recognized character outcomes are obtained. In a software environment, the novel model is implemented; also, for exhibiting the proposed technique’s superiority, the outcomes are analogized with the prevailing approaches.