A Novel Biologically Inspired Visual Cognition Model: Automatic Extraction of Semantics, Formation of Integrated Concepts, and Reselection Features for Ambiguity
Peijie Yin,Hong Qiao,Wei Wu,Lu Qi,Yinlin Li,Shanlin Zhong,Bo Zhang
DOI: https://doi.org/10.1109/tcds.2017.2749978
IF: 4.546
2018-01-01
IEEE Transactions on Cognitive and Developmental Systems
Abstract:Techniques that integrate neuroscience and information science benefit both fields. Many related models have been proposed in computer vision; however, in general, the robustness and recognition precision are still key problems in object recognition models. In this paper, inspired by the process by which humans recognize objects and its biological mechanisms, a new integrated and dynamic framework is proposed that mimics the semantic extraction, concept formation and feature reselection found in human visual processing. The main contributions of the proposed model are as follows: 1) semantic feature extraction: local semantic features are learned from episodic features extracted from raw images using a deep neural network; 2) integrated concept formation: concepts are formed using the local semantic information and structural information is learned through a network; and 3) feature reselection: when ambiguity is detected during the recognition process, distinctive features based on the differences between the ambiguous candidates are reselected for recognition. Experimental results on four datasets show that-compared with other methods-the new proposed model is more robust and achieves higher precision for visual recognition, especially when the input samples are semantically ambiguous. Meanwhile, the introduced biological mechanisms further strengthen the interaction between neuroscience and information science.