Face Gesture Change Detection Using Invariant Baseline Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately detecting changes in face gestures due to factors like head pose, lighting, and motion effects, requiring large training datasets and manual user intervention, which affect computational efficiency and accuracy.
Innovation Solution
A biometric system is used to extract features from images that are invariant to head pose, lighting, and motion effects, allowing for unsupervised analysis of a sequence of images to compute a baseline representation and detect changes in face gestures without labeled training datasets, using a two-step approach to filter images for significant changes before applying machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing methods are used to detect face gestures, then the system can capture detailed facial expressions, but it becomes highly sensitive to environmental variations like lighting, head pose, and motion effects
Solution Approach 1:
The patent extracts only the essential facial features that are invariant to environmental variations. By selecting specific facial landmarks and geometric relationships that remain stable under different lighting, pose, and motion conditions, the system filters out harmful environmental factors while preserving the core gesture information.
Solution Approach 2:
The patent transforms raw pixel data into normalized geometric parameters such as distances between facial landmarks, angles, and ratios. These parameter transformations make the detection system invariant to lighting changes, head pose variations, and motion effects, as the geometric relationships remain consistent despite environmental changes.
2Measurement precision
If large training datasets with manual labeling are used, then the machine learning model achieves higher accuracy, but the computational cost and time requirements increase significantly
Solution Approach 1:
The patent performs preliminary feature extraction and baseline computation before the actual detection process. By pre-processing the images to extract invariant facial features and establishing baseline representations, the system reduces the complexity of the subsequent detection task, eliminating the need for extensive training datasets and reducing computational requirements.
Solution Approach 2:
The system uses the subject's own baseline facial representation as the reference for detection, rather than requiring external training datasets. The baseline is automatically computed from the subject's images themselves, making the system self-sufficient and eliminating the need for manual labeling and large-scale training data collection.
3Speed
If real-time processing is implemented for face gesture detection, then the system responds quickly to changes, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent segments the detection process into distinct stages: baseline computation, feature extraction, representation computation, and comparison. This segmentation allows each stage to be optimized independently, with feature extraction using efficient algorithms and comparison using simple distance metrics, enabling real-time processing without excessive computational complexity.
Solution Approach 2:
The patent extracts only the necessary facial features required for gesture detection rather than processing the entire image. By focusing computation on specific facial landmarks and their geometric relationships, the system achieves real-time performance with reduced computational complexity, avoiding unnecessary processing of irrelevant image regions.
Data Source
AI summary
There is provided a computer implemented method of analyzing a sequence of images for detection of a change in face gesture, comprising: analyzing the sequence of time spaced images depicting a face of a subject to compute a baseline representation denoting a baseline state of the subject, for a respective image of the sequence of time spaced images depicting a face of a subject: extracting a plurality of features from the face of the subject of the respective image, computing a respective representation from the plurality of features, comparing the respective representation to the baseline representation, and detecting a change in face gesture of the subject when the comparison is above a threshold.