A New Fusion Feature Selection Model (FFSM) based Feature Extraction System for Hand Gesture Recognition
The research presented here proposes a unique approach for exact feature extraction that makes use of the Fusion Feature Selection Model (FFSM). The method encompasses various stages to preprocess and extract crucial features from first-person hand action (FHPA) images. Preprocessing involves video-to-frame conversion, RGB to grayscale conversion, an improved median filter, and Gaussian blur-based image smoothing. Segmentation is achieved using the Improved SwinNet to identify meaningful regions within the images. Feature extraction employs the Gabor Line Derivative (GLD) method, Active Shape Model (ASM), and Histogram of Oriented Gradients (HOG) to capture texture, edge, and shape information, respectively. Extensive experimental evaluations demonstrate the effectiveness of our proposed approach, achieving remarkable performance in accurate feature extraction tasks.