Gesture Trajectory Recognition for Following Robots
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hand trajectory recognition methods for following robots are limited by the need for extensive training data, sensitivity to gesture order, and high computational requirements, making them inflexible and resource-intensive, while also being sensitive to translation, scaling, and rotation.
Innovation Solution
A hand trajectory recognition method using a kinect camera to sample and smooth hand movement trajectories, recording velocity and shape descriptors, and calculating cosine similarities to recognize gestures independently of order, translation, scaling, and rotation, with a comprehensive similarity criterion for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional hand trajectory recognition methods are used, then gesture recognition accuracy can be achieved, but extensive training data is required and computational complexity is high
Solution Approach 1:
The patent segments the hand trajectory into velocity descriptor and shape descriptor components. The velocity descriptor captures movement dynamics while the shape descriptor captures geometric characteristics. This segmentation allows independent analysis of different trajectory aspects, reducing the need for extensive training data while maintaining recognition accuracy.
Solution Approach 2:
The patent extracts key features from the hand trajectory by computing velocity vectors and shape descriptors. By taking out only the essential characteristics (velocity magnitude/direction and shape geometry) rather than processing the entire trajectory raw data, the computational complexity is significantly reduced while preserving recognition accuracy.
2Reliability
If traditional gesture recognition methods are used, then recognition can be performed, but the gesture must be completed in a set order which reduces flexibility
Solution Approach 1:
The patent transforms the trajectory recognition from time-order dependent to shape-space dependent by using shape descriptors. Instead of comparing trajectories point-by-point in temporal sequence, the method compares geometric shapes in a transformed space, allowing gestures to be recognized regardless of the order in which they are performed, thus increasing flexibility while maintaining reliability.
3Measurement precision
If detailed trajectory analysis is performed, then recognition accuracy improves, but resource consumption increases
Solution Approach 1:
The patent extracts only the essential features from the complete trajectory data - specifically velocity descriptors (magnitude and direction) and shape descriptors. By extracting these key characteristics rather than performing detailed analysis of all trajectory points, the method achieves good recognition accuracy with reduced computational resource consumption.
Solution Approach 2:
The patent applies different analysis methods to different aspects of the trajectory: velocity analysis for movement dynamics and shape analysis for geometric characteristics. This localized approach allows efficient processing of each aspect using methods optimized for that specific property, reducing overall computational resources while maintaining accuracy.
4Reliability
If conventional trajectory recognition is used, then gestures can be recognized, but the system is sensitive to translation, scaling, and rotation
Solution Approach 1:
The patent transforms the trajectory representation into shape descriptors that are inherently invariant to translation, scaling, and rotation. By representing trajectories as normalized geometric shapes rather than absolute coordinate sequences, the system becomes insensitive to these transformations, improving reliability in varied operating conditions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A hand trajectory recognition method for a following robot based on hand velocity and trajectory distribution comprises: sampling and photographing an operator by a kinect camera to obtain hand projection plane data; carrying out moving average and smooth processing on the hand projection plane data, establishing velocity vectors, and processing the velocity vectors to obtain a hand movement descriptor; establishing a hand movement area, traversing hand three-dimensional positions of all frames in an order of sampling and photographing, assigning a mesh where the hand three-dimensional position of each frame is located, and calculating centroid positions of all assigned meshes; establishing centroid directing vectors, and processing the centroid directing vectors to obtain a hand trajectory shape descriptor; and processing cosine values of two angles to obtain a common similarity of the movement descriptor and the trajectory shape descriptor to standard descriptor, and using a standard gesture with a maximum common similarity as a result. The invention can accurately recognize the class of a gesture, is insensitive to the translation, scaling, rotation and order of hand trajectories, is high in flexibility and can save time and energy.