Dynamic Gesture Recognition via Finger Joint Angular Change

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing human gesture recognition methods can only recognize static gestures and are unable to detect dynamic and orderly gesture movements, as they rely on recognizing a single gesture image and do not account for angular changes of finger joints over time.

Innovation Solution

A method that recognizes gesture movements by calculating the total angular change of finger joints between two gesture images using a mathematical model, determining if a user has performed a specified gesture based on the angular changes and a preset threshold, allowing for the recognition of dynamic and orderly gestures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single gesture image is recognized using a convolutional neural network, then the recognition of static gestures is achieved, but the recognition of dynamic gesture movements is not possible

Engineering Contradiction:
Improvegesture recognition accuracyVSAvoidgesture type coverage
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transitions from static gesture recognition to dynamic gesture movement recognition by introducing temporal dimension. It captures multiple gesture images at different time points and calculates angular changes of finger joints between these images, enabling the system to recognize not only static gestures but also dynamic gesture movements and their sequences.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent adds the time dimension to the traditional two-dimensional gesture image recognition. By capturing gesture images at different time points and calculating angular changes over time, it transforms the recognition task from spatial-only to spatio-temporal, enabling detection of gesture movements and sequences.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If only static gesture images are recognized, then the system complexity is low, but the application scope in real-time scenarios is limited

Engineering Contradiction:
Improveapplication scenario rangeVSAvoidrecognition system complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the gesture recognition task into distinct components: capturing multiple gesture images at different time points, detecting finger joint positions in each image, calculating angular changes between consecutive images, and comparing against threshold values. This segmentation makes the dynamic recognition system more manageable and implementable.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces complex mechanical or hardware-based gesture capture systems with a software-based image processing approach. By using convolutional neural networks to detect finger joint positions from standard images and calculating angular changes through mathematical operations, it achieves dynamic gesture recognition without requiring specialized hardware.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Adaptability or versatility

If angular changes of finger joints are calculated between multiple gesture images, then dynamic gesture movements can be recognized, but the processing time and computational complexity increase

Engineering Contradiction:
Improvegesture movement recognition capabilityVSAvoidprocessing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent focuses on calculating angular changes only for specific finger joints that are relevant to the gesture being recognized, rather than processing all possible joint angles in all images. This partial action approach reduces computational overhead while maintaining recognition accuracy for the target gestures.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent transforms the raw image data into angular change parameters between consecutive frames, which are then compared against threshold values. By changing the parameter representation from pixel-based image data to angular difference values, the system enables faster comparison and decision-making for gesture recognition.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3805982B1Gesture recognition method, apparatus and device
Publication Date: 2023.10.25 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3805982B1 patent drawingFigure 1
  • EP3805982B1 patent drawingFigure 2~3
  • EP3805982B1 patent drawingFigure 4

AI summary

Provided is a gesture recognition method. The method obtains a first vector and a second vector representing an angle of a finger joint by means of performing recognition on a first gesture image and a second gesture image, and then uses a mathematical model method to calculate and obtain a first total angle change according to the first vector and the second vector. The first total angle change is the total angle change corresponding to a first specific joint of a first specific finger. A gesture recognition result is obtained according to the first total angle change and a first predetermined threshold value. Since there must be angle change in a specific joint of a specific finger when a user performs a specific gesture, the method can recognize a motion gesture according to changes in the angle of the specific joint of the specific finger in two gesture images. The gesture recognition method has broad application in the field of artificial intelligence. A gesture recognition apparatus is also provided.