A Review on Vision-based Hand Gesture Recognition Targeting RGB-Depth Sensors

With the advancement of automation, vision-based hand gesture recognition (HGR) is gaining popularity due to its numerous uses and ability to easily communicate with machines. However, identifying hand positions is the most difficult assignment due to the fact of crowded backgrounds, sensitivity to light, form, speed, size, and self-occlusion. This review summarizes the most recent studies on hand postures and motion tracking using a vision-based approach by applying Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA). The parts and subsections of this review article are organized into numerous categories, the most essential of which are picture acquisition, preprocessing, tracking and segmentation, feature extraction, collation of key gesture identification phases, and classification. At each level, the various algorithms are evaluated based on critical key points such as localization, largest blob, per pixel binary segmentation, depth information, and so on. Furthermore, the datasets and future scopes of HGR approaches are discussed considering merits, limitations, and challenges.