“School of Cognitive Sciences”
Back to Papers HomeBack to Papers of School of Cognitive Sciences
| Paper IPM / Cognitive Sciences / 11352 |
|
||||
| Abstract: | |||||
|
This paper introduces a novel, effective applicability of features inspired by visual ventral stream and biologically-motivated classification model, mixture of experts network for face/nonface recognition task. It describes a feature extracting system that derives from a feedforward model of visual cortex and builds a set of pose, facial expression, illumination and view invariant C1 features from all images in the dataset. Also, mixture of MLP experts network is a classifier which demonstrates high generalization capabilities in many different tasks. In accordance to these biological evidences, we propose face/nonface recognition model which combine these two techniques for the robust face/nonface problem. Experimental results using the combination C1 features and mixture of MLP experts network classifier, obtains higher recognition rate than related works in face/nonface identification. In addition, experimental results demonstrate this method is illumination and view-invariant.
Download TeX format |
|||||
| back to top | |||||
