Application of a New Multiple Class Measurement Function in Face Recognition
Nearest subspace and collaborative representation residuals have strong underlying geometric meaning. They get outstanding results in multiple class classification problems. Nearest sub-space classifier uses a local measurement, considering relationship between testing sample and each class each time. Collaborative representation classifier uses global measurement which considers both intra-class and inter-class measurements. These two measurements are inde-pendent. By introducing product measurement, they are combined for the multiple class classi-fication problems. In face recognition problem, product measurement classifier is compared with the-state-of the-art algorithms and outperforms the other methods.
Multiple Class ProblemProduct MeasurementFace Recognition
NETL
NSTL
万方数据
维普
