Hyperparameter Tuning of Deep Convolutional Neural Network for Hand Gesture Recognition
S. Padmakala, Saif O. Husain, Ediga Poornima, Papiya Dutta, Mukesh Soni · 2024
Hand Gesture Recognition (HGR) is a useful application of Human Computer Interaction (HCI) but that still faces difficulty in obtaining high recognition speed and accuracy. Deep Learning (DL) has the ability to quickly and precisely pick up the features of a complex object while demonstrating exceptional performance in speech, face recognition etc. In order to address the challenge of HGR in intricate scenes, a novel method named Adaptive Habitat Biogeography-Based Optimizer (AHBBO) for tuning the hyperparameters of a Deep Convolutional Neural Network (DCNN) model. In the present research, Hand Gesture Recognition Image Dataset (HaGRID) is used which comprises gesture labels to solve hand gesture classification. Compared traditional approaches such as Multi-Culture Sign Language Using Graph and General DL Network (MCSL-GGDLN), Faster Regional Convolutional Neural Network (Faster-RCNN), and Adaptive Hough Transform with DCNN (AHT-DCNN), the AHBBO-DCNN achieves accuracy of $99.53 \%$ and demonstrated that it is able to detect the gesture recognition categories accurately.