Deep-learning-based reading eye-movement analysis for aiding biometric recognition

Xiaoming Wang, Xinbo Zhao, Yanning Zhang

科研成果: 期刊稿件文章同行评审

10 引用 (Scopus)

摘要

Eye-movement recognition is a new type of biometric recognition technology. Without considering the characteristics of the stimuli, the existing eye-movement recognition technology is based on eye-movement trajectory similarity measurements and uses more eye-movement features. Related studies on reading psychology have shown that when reading text, human eye-movements are different between individuals yet stable for a given individual. This paper proposes a type of technology for aiding biometric recognition based on reading eye-movement. By introducing a deep-learning framework, a computational model for reading eye-movement recognition (REMR) was constructed. The model takes the text, fixation, and text-based linguistic feature sequences as inputs and identifies a human subject by measuring the similarity distance between the predicted fixation sequence and the actual one (to be identified). The experimental results show that the fixation sequence similarity recognition algorithm obtained an equal error rate of 19.4% on the test set, and the model obtained an 86.5% Rank-1 recognition rate on the test set.

源语言英语
页(从-至)390-398
页数9
期刊Neurocomputing
444
DOI
出版状态已出版 - 15 7月 2021

指纹

探究 'Deep-learning-based reading eye-movement analysis for aiding biometric recognition' 的科研主题。它们共同构成独一无二的指纹。

引用此