3D Gesture Recognition via Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gesture recognition technologies face challenges in accurately detecting symbols drawn in three-dimensional space, particularly due to the lack of defined start or end points and the inability to select the correct portion of a continuous movement, leading to misclassification of user inputs.
Innovation Solution
The use of multiple neural networks trained for different symbols allows for the prediction of complete symbols from partial movements, enabling the selection of the appropriate portion of the object path and accounting for distortions, thereby improving the accuracy of gesture recognition in three-dimensional space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional gesture recognition methods are used, then the system is simple to implement, but the accuracy of detecting symbols in three-dimensional space is poor
Solution Approach 1:
The patent divides the continuous object path into multiple segments and uses multiple neural networks, each trained to recognize specific symbols. Each neural network processes a segment of the path independently to determine the complete symbol, allowing accurate recognition without requiring the entire continuous movement to be processed as a single unit.
Solution Approach 2:
The patent transitions from two-dimensional touch display recognition to three-dimensional space recognition by using sensors that capture object movement in three-dimensional space. This dimensional change enables the system to detect gestures made with objects like styluses or fingers in mid-air, significantly improving gesture recognition accuracy.
2Loss of time
If the system waits for defined start and end points, then the gesture detection is precise, but the response time increases
Solution Approach 1:
The patent uses neural networks to predict the complete symbol based on partial object paths. As the object moves through three-dimensional space, the system continuously processes segments of the path and predicts the intended symbol in advance, rather than waiting for the object to complete its movement. This preliminary prediction reduces detection delay while maintaining accuracy.
Solution Approach 2:
The system provides feedback by comparing the predicted complete symbol with the actual object path segments. This feedback mechanism allows the system to adjust its predictions in real-time, ensuring accurate gesture recognition even when the object movement is continuous and without clearly defined boundaries.
3Measurement precision
If multiple neural networks are used to recognize different symbols, then the recognition accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent segments the recognition task by dividing the object path into multiple segments and assigning different neural networks to process different segments. Each neural network is specialized for recognizing specific symbols, allowing the system to parallelize the recognition process and reduce computational complexity while maintaining high accuracy.
Solution Approach 2:
The system uses partial action by processing only the necessary segments of the object path with the appropriate neural networks. Instead of processing the entire continuous path with all neural networks simultaneously, the system selectively processes segments, reducing computational load while ensuring accurate symbol recognition.
Data Source
AI summary
The present disclosure relates generally to computational devices and gesture recognition. A computational device may include instructions that determine, based on a parameterization for a portion of a symbol, complete predicted symbols. The complete predicted symbols may be used as an input for an application operating on the computational device.