Human Expression Recognition Using Compressed Sensing Algorithm

In this paper, we present a novel facial expression recognition algorithm based on compressed sensing principles utilizing Gabor features. The implementation involves extracting multi-scale and multi-orientation Gabor features from facial images to capture localized texture information. Our compressed sensing approach employs sparse representation techniques where facial expressions are reconstructed from significantly fewer measurements than traditional methods require. We implemented and tested this algorithm on the JAFFE database, with experiments demonstrating high recognition accuracy and robust performance across various expression categories. The core algorithm implementation includes feature dimension reduction through random projection matrices and sparse coding using optimization techniques like L1-norm minimization. By leveraging compressed sensing technology, we effectively reduce data processing and storage costs while maintaining competitive algorithm performance, thereby providing new approaches for practical facial expression recognition applications. The technical implementation involves constructing an overcomplete dictionary of expression prototypes and solving the sparse recovery problem using algorithms such as Orthogonal Matching Pursuit (OMP) or Basis Pursuit. Future work will focus on algorithm refinement and optimization to develop more accurate and efficient facial expression recognition systems, potentially incorporating deep learning components for enhanced feature learning.