Vehicle Hand Gesture Recognition Using Normalized Key Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hand gesture recognition systems for vehicles require significant computing power and large amounts of training data, making them difficult to deploy on vehicles with lower power electronic control units and increasing the time for training and recognition, while also being sensitive to hand size and position in images.
Innovation Solution
A method and system that detect hand key points, including a wrist joint and finger joint points, normalize their coordinates relative to the wrist joint, and feed these normalized coordinates into a hand gesture classification model for recognition, reducing the need for extensive training data and computing power, and enabling position-angle invariance for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If computer vision technology is used to achieve hand gesture recognition by inputting entire images for classification, then recognition accuracy can be maintained, but computing power requirements increase significantly
Solution Approach 1:
The patent extracts only the essential hand key point coordinates from the entire image, rather than processing the full image. By identifying and extracting coordinates of key hand joints (wrist, finger joints, thumb joints), the system reduces the input data dimensionality while maintaining recognition accuracy, thus lowering computing power requirements
Solution Approach 2:
The patent segments the hand gesture recognition process into distinct components: hand region detection, key point coordinate extraction, coordinate normalization, and gesture classification. This segmentation allows each component to be optimized independently, with the coordinate extraction step significantly reducing computational load compared to processing entire images
2Measurement precision
If a large amount of image data is used for model training, then model accuracy can be improved, but training time increases significantly
Solution Approach 1:
The patent extracts only the necessary hand key point coordinates as training features, rather than using entire images as training samples. This extraction approach dramatically reduces the data dimensionality and complexity, enabling faster model training while maintaining the ability to achieve high recognition accuracy with fewer training samples
Solution Approach 2:
The patent changes the parameter representation from full image pixels to normalized hand key point coordinates. This parameter transformation reduces the complexity of training data, allowing the model to learn effective gesture patterns with fewer samples and less training time
3Loss of information
If hand gesture recognition is performed using traditional methods, then comprehensive image information is processed, but the system becomes sensitive to hand size and position in the image
Solution Approach 1:
The patent applies coordinate normalization that transforms hand key point coordinates into a standardized reference system where the wrist joint is the origin and distances are normalized by the distance from wrist to the farthest finger joint. This parameter transformation makes the representation invariant to hand size, position, and scale, while preserving the essential geometric relationships needed for accurate gesture recognition
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The invention provides a hand gesture recognition method, system and medium for a vehicle, the method comprising: obtaining an input image to be recognized; detecting a hand region in the input image; extracting coordinates of hand key points in the hand region, wherein the hand key points include a wrist j oint point and a plurality of finger j oint points; normalizing the coordinates of each of the plurality of finger j oint points relative to the wrist j oint point by utilizing a distance between the wrist j oint point and an adjacent finger j oint point with the wrist joint point as an origin; and feeding the normalized coordinates of each of the plurality of finger joint points relative to the wrist joint point into a hand gesture classification model to recognize a hand gesture. The invention also provides a vehicle with the above described hand gesture recognition function.