Hand Gesture Recognition Using HOG Descriptors and Skin Color Filtering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for recognizing hand gestures in complex backgrounds or with overlapping skin color regions are inefficient and unreliable, requiring large datasets and being vulnerable to lighting variations and noise.

Innovation Solution

The method involves scanning images using a window to generate HOG descriptors, measuring resemblance values between scanned and query templates, and judging gestures based on predetermined conditions, employing HSI color models and HOS maps to improve accuracy and robustness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If skin color region detection is used in complex backgrounds, then hand detection capability is improved, but false detection increases when other skin color regions exist

Engineering Contradiction:

Improvehand detection accuracyVSAvoidfalse detection rate

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent divides the hand detection process into multiple stages: skin color region detection, candidate region extraction, HOG feature extraction, and template matching. This segmentation allows each stage to focus on specific characteristics, improving overall accuracy while reducing false detections from skin-colored objects

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different detection strategies to different regions: skin color detection is used to identify potential hand regions, while HOG descriptors and template matching are applied specifically to these candidate regions. This local quality approach ensures high accuracy for actual hands while filtering out false skin-colored regions

Inventive Principle:

Principle #3Local quality

2Reliability

If Gaussian mixture model learning is conducted with many data sets, then model reliability is improved, but data processing time and complexity increase

Engineering Contradiction:

Improvemodel reliabilityVSAvoiddata processing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent performs Gaussian mixture model learning in advance to create a pre-trained skin color model. This preliminary action allows the system to quickly apply the model to new images without requiring extensive real-time data processing, thus maintaining reliability while reducing processing time

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent extracts only the essential skin color characteristics using the pre-trained Gaussian mixture model, rather than processing all possible features. This extraction approach maintains model reliability while significantly reducing the data processing burden during actual hand detection

Inventive Principle:

Principle #2Taking out (Extraction)

3Ease of operation

If hand movement detection uses brightness difference between frames, then movement detection is enabled, but detection accuracy decreases due to camera noise

Engineering Contradiction:

Improvemovement detection capabilityVSAvoidhand movement detection accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent introduces skin color detection as an intermediary step between frame differencing and hand identification. The skin color model acts as a mediator that filters out noise-induced false movements while preserving actual hand movements, thereby improving detection accuracy without losing movement detection capability

Inventive Principle:

Principle #24Intermediary (Mediator)

4Productivity

If integral image characteristic is used for hand candidate detection, then detection speed is improved, but detection accuracy decreases in complex backgrounds

Engineering Contradiction:

Improvedetection speedVSAvoidhand candidate detection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent merges integral image characteristics with HOG descriptors and template matching. The integral image provides fast initial candidate detection, while HOG features and template matching are applied to these candidates to improve accuracy in complex backgrounds, combining the speed advantage with the precision advantage

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS8649559B2Method and interface of recognizing user's dynamic organ gesture and electric-using apparatus using the interface

Publication Date: 2014.02.11 LG DISPLAY CO LTD

AI summary

A method of recognizing a user's dynamic organ for use in an electric-using apparatus includes scanning a target image inputted through an imaging element using a window; generating a HOG descriptor of a region of the target image that is scanned when it is judged that the scanned region includes a dynamic organ; measuring a resemblance value between the HOG descriptor of the scanned region and a HOG descriptor of a query template for a gesture of the dynamic organ; and judging that the scanned region includes the gesture of the dynamic organ when the resemblance value meets a predetermined condition.