Facial Analysis Image Normalization via Segmented AU Feature Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial analysis is hindered by variations in image orientation and pose, making consistent analysis of facial images challenging due to differences in facial feature positions and distortions.
Innovation Solution
A method involving the selection of base facial features from a neutral expression image, which are used to normalize analysis images through techniques like Procrustes analysis, ensuring consistent alignment and feature emphasis based on the specific Action Units (AUs) being detected, facilitating accurate prediction of AU presence and intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single base set of facial features is used to normalize all analysis images, then the normalization process is simple and fast, but the accuracy of detecting different Action Units (AUs) deteriorates due to varied facial expressions and poses
Solution Approach 1:
The patent segments the facial feature set into multiple base sets, where each base set is specifically associated with certain Action Units (AUs). Instead of using a single universal base feature set, the system divides features into specialized groups (e.g., one base set for AU 1 detection, another for AU 2 detection, etc.), allowing each normalization process to focus on the relevant features for that specific AU, thereby improving detection accuracy while maintaining efficient processing
Solution Approach 2:
The patent implements a dynamic selection mechanism where the appropriate base feature set is automatically selected based on which AU is being detected. The system adapts the normalization process by choosing different base feature sets dynamically according to the detection task, rather than using a static single base set for all scenarios. This dynamic adaptation optimizes the normalization for each specific AU detection task
2Measurement precision
If multiple specialized base sets of facial features are used for different Action Units, then the accuracy of AU detection is improved, but the complexity of the normalization system increases
Solution Approach 1:
The patent creates a universal normalization framework that can handle multiple different AUs through a common system architecture. The system uses a single normalization module that can accept different base feature sets as input, making the system multi-functional. This universal approach allows the same normalization infrastructure to serve multiple AU detection purposes without requiring separate dedicated systems for each AU, thereby managing complexity while maintaining specialized accuracy
Solution Approach 2:
The patent performs preliminary organization of facial features into multiple pre-defined base sets during system initialization or preprocessing. Each base set is prepared in advance and associated with specific AUs. This preliminary action eliminates the need for complex real-time feature organization during detection, as the system only needs to select from pre-prepared base sets, thereby reducing operational complexity while maintaining detection accuracy
3Measurement precision
If multiple specialized base sets of facial features are used for different Action Units, then the precision of facial analysis is improved, but the computational time and resources increase
Solution Approach 1:
The patent applies local quality optimization by creating base feature sets that are locally specialized for specific AUs. Each base feature set contains features that are locally relevant to particular facial regions and AUs (e.g., eye-related features for AU 1, mouth-related features for AU 12). This local specialization means that only the necessary features for each AU are processed in detail, rather than processing all facial features uniformly, thereby reducing computational overhead while maintaining high precision for each specific AU detection
Data Source
AI summary
A method may include obtaining a base facial image, and obtaining a first set of base facial features within the base facial image, the first set of base facial features associated with a first facial AU to be detected in an analysis facial image. The method may also include obtaining a second set of base facial features within the base facial image, the second set of facial features associated with a second facial AU to be detected. The method may include obtaining the analysis facial image, and applying a first image normalization to the analysis facial image using the first set of base facial features to facilitate prediction of a probability of the first facial AU. The method may include applying a second image normalization to the analysis facial image using the second set of base facial features to facilitate prediction of a probability of the second facial AU.