Hand Gesture Detection Using Normalized Keypoints and Decision Trees

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing computer vision-based hand gesture detection technologies are limited in the types of gestures recognized, result in less intuitive user interactions, suffer from high false positives, and are computationally complex, making them impractical for devices with limited resources.

Innovation Solution

A decision tree-based approach for hand gesture classification that normalizes hand keypoint locations and uses a trained decision tree ensemble to reduce false positives, along with depth data to differentiate between touching and hovering gestures, without relying on deep neural networks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep neural networks are used for computer vision-based gesture detection, then gesture detection accuracy is improved, but computational complexity and resource requirements increase significantly

Engineering Contradiction:

Improvegesture detection accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the gesture detection task into multiple components: hand detection, keypoint extraction, gesture classification, and validation. Each component is processed separately through different neural networks and algorithmic steps, allowing optimized resource allocation for each subtask while maintaining overall accuracy

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs partial gesture recognition by focusing on detecting specific hand keypoints and essential gesture features rather than complete hand pose estimation. This selective approach reduces computational load while maintaining sufficient accuracy for gesture control applications

Inventive Principle:

Principle #16Partial or excessive action

2Measurement precision

If deep neural networks are used for computer vision-based gesture detection, then gesture detection accuracy is improved, but processing resources and memory requirements increase

Engineering Contradiction:

Improvegesture detection accuracyVSAvoidprocessing resources

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system extracts only the essential features needed for gesture detection: hand keypoint locations and basic gesture classification labels. By taking out only the critical information rather than processing complete hand pose data, the system reduces memory requirements and processing resources while maintaining detection accuracy

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system uses lightweight neural network models that can be quickly instantiated and discarded for each frame processing. These temporary, resource-efficient models perform the necessary detection tasks without requiring persistent high-resource allocations, enabling smooth operation on mobile devices with limited processing power

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

3Adaptability or versatility

If computer vision-based gesture detection is used, then user interaction with non-digital content is enabled, but false positive detection rate increases

Engineering Contradiction:

Improveuser interaction capabilityVSAvoidfalse positive rate

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system implements feedback mechanisms where detected gestures are validated against expected gesture patterns and user interaction context. Invalid or unexpected gestures are rejected or flagged for re-detection, reducing false positives while maintaining the system's ability to recognize diverse valid gestures

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary hand detection and keypoint extraction before attempting gesture classification. This preliminary action filters out non-hand objects and invalid input regions, preventing false positive gesture detections from non-hand elements in the scene

Inventive Principle:

Principle #10Preliminary action

4Use of energy by moving object

If limited gesture types are recognized, then computational resources are reduced, but user interaction intuitiveness decreases

Engineering Contradiction:

Improvecomputational resourcesVSAvoiduser interaction intuitiveness

Core Design Contradiction:

Use of energy by moving objectVSEase of operation

Solution Approach 1:

The system implements universal gesture recognition that can identify multiple gesture types (pointing, pinching, grasping, open hand) using the same core detection pipeline. This multi-functionality allows diverse intuitive interactions while maintaining resource efficiency through a unified processing approach rather than separate specialized detectors for each gesture

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS12424029B2Devices and methods for single or multi-user gesture detection using computer vision

Publication Date: 2025.09.23 HUAWEI TECH CO LTD

AI summary

Methods and devices are described for computer vision-based gesture detection. From a frame of image data, extracted locations of keypoints of a detected hand are obtained. The extracted locations are normalized to obtain normalized features. The normalized features are processed using a trained decision tree ensemble to generate a probability of a valid gesture for the detected hand. The generated probability is compared with a defined decision threshold to generate a binary classification to classify the detected hand as a valid gesture or invalid gesture.