Gabor Filter and Random Forest for Facial Expression Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current facial-expression recognition algorithms, especially those using machine learning and deep learning, face challenges in achieving real-time recognition due to complex algorithm structures and high computational requirements, leading to low accuracy on terminal devices like ARM development boards, and existing methods like Active Shape Model and deep convolution networks suffer from feature errors and high complexity.
Innovation Solution
An image processing method that combines a Gabor filter with a random-forest classifier, where the Gabor filter extracts a characteristic vector from face images using a reduced dimension and direction filter bank, and the random-forest classifier is trained using K-fold cross-validation with grading values to determine facial expressions, allowing for real-time facial-expression recognition with high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning and deep learning algorithms are used for facial-expression recognition, then recognition accuracy can be improved, but algorithm complexity and computational requirements increase, making real-time recognition difficult on terminal devices
Solution Approach 1:
The patent segments the facial-expression recognition task into distinct components: feature extraction using Gabor filters, dimensionality reduction, and classification using random-forest classifier. This segmentation allows each component to be optimized independently, reducing overall algorithm complexity while maintaining accuracy
Solution Approach 2:
The patent replaces complex, resource-intensive deep learning models with lighter-weight algorithms (Gabor filters and random-forest classifier) that consume fewer computational resources. This substitution enables real-time processing on terminal devices with limited hardware capabilities
2Measurement precision
If complex algorithms like deep convolution networks are used, then feature extraction capability is improved, but processing time increases and real-time recognition on ARM development boards becomes difficult
Solution Approach 1:
The patent extracts only the essential features from facial images using Gabor filters, focusing on key texture and pattern information. This selective extraction eliminates unnecessary computational steps present in full deep convolution networks, significantly reducing processing time while maintaining feature extraction accuracy
Solution Approach 2:
The patent changes the parameters of the filtering and classification processes by using optimized Gabor filter configurations and random-forest parameters tailored for rapid processing. This parameter optimization enables the system to achieve accurate feature extraction with minimal processing time on resource-constrained devices
3Device complexity
If traditional algorithms like Active Shape Model are used, then computational requirements are reduced, but feature errors increase and recognition accuracy decreases
Solution Approach 1:
The patent creates a composite approach by combining Gabor filter-based feature extraction with random-forest classification. This composite method leverages the strengths of both traditional signal processing and modern machine learning, achieving high accuracy without the excessive computational requirements of pure deep learning approaches
Data Source
AI summary
An image processing method, an image processing device, a training method and a computer-readable storage medium. The image processing method includes: extracting a characteristic vector in an image to be recognized; based on the characteristic vector of the image to be recognized, acquiring a predicted score value of the image to be recognized; and based on the predicted score value, determining a category of an image information of the image to be recognized; wherein the image to be recognized is a face image, and the image information is a facial expression.