Hand Segmentation Using Depth Range Data for Gesture Analysis

The patent applies segmentation by dividing the hand region into multiple depth ranges (e.g., first depth range for palm, second depth range for fingers) and processing each segment separately. This allows the system to maintain high accuracy through detailed depth-based segmentation while reducing overall processing time by handling smaller, more manageable regions independently rather than processing the entire hand as one large region.

The patent extracts and removes the forearm region from the target region, retaining only the hand portion for gesture analysis. By taking out the forearm (which has different geometric characteristics and movement patterns), the system maintains comprehensive gesture detection coverage for hand gestures while significantly reducing classification complexity, as hand gestures have more distinct and easier-to-recognize geometric features compared to arm movements.

The patent leverages depth information as an additional dimension to enhance segmentation accuracy. By incorporating depth range data into the segmentation process, the system can distinguish between different parts of the hand (palm, fingers, knuckles) based on their spatial depth characteristics, achieving high accuracy without significantly sacrificing processing speed, as the depth data provides direct geometric cues that simplify the segmentation task.