Face Shape-Based Salient Region Extraction for Expression Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional facial expression recognition methods fail to accurately account for differences in face shapes, leading to unreliable feature extraction and decreased recognition accuracy, especially when using 2D images.
Innovation Solution
An apparatus that detects feature points, determines face shape, and identifies pre-defined shape groups to extract features from 'salient regions' specific to each face shape, utilizing both 2D and 3D image information to enhance recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the same regions are used for feature extraction regardless of face shape, then the processing method is simple, but the recognition accuracy decreases
Solution Approach 1:
The patent applies local quality by determining different salient regions based on face shape characteristics. Instead of using uniform regions for all faces, the system identifies shape-specific regions (e.g., oval, round, square faces have different salient areas) and extracts features only from those personalized regions, thereby improving recognition accuracy while maintaining reasonable processing complexity through automated shape classification
Solution Approach 2:
The patent segments the face into shape-specific salient regions based on detected face shape. The face region is divided into multiple candidate regions, and only those matching the detected face shape are selected as salient regions for feature extraction. This segmentation approach allows the system to focus computational resources on relevant areas, improving accuracy without proportionally increasing overall complexity
2Measurement precision
If face shape-specific regions are used for feature extraction, then recognition accuracy increases, but the system complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-defining multiple candidate regions and their corresponding face shapes before actual recognition. During operation, the system only needs to detect the face shape and select from pre-prepared region configurations, rather than dynamically computing optimal regions. This preliminary preparation significantly reduces real-time computational complexity while maintaining high recognition accuracy
Solution Approach 2:
The patent changes the parameter of region selection based on face shape parameters. Different face shapes (oval, round, square, etc.) correspond to different sets of salient regions. The system adjusts which regions are activated for feature extraction according to the detected face shape parameter, enabling adaptive high-accuracy recognition without requiring a completely complex system architecture
Data Source
AI summary
One of the aspects of the present disclosure discloses an apparatus for recognizing expression of a face in a face region of an image, comprising: a unit configured to detect feature points of the face in the face region of the image; a unit configured to determine a face shape of the face in the face region based on the detected feature points; a unit configured to determine a group that the face in the face region belongs to based on the determined face shape and pre-defined shape groups; and a unit configured to determine an expression of the face in the face region based on pre-generated first models corresponding to the determined group and features extracted from at least one region in the face region, wherein the at least one of the regions are regions which are labeled out in the pre-generated first models corresponding to the determined group.