Machine learning methods to analyze subliminal priming ERP

• Main Author: Wu, Zuobin
• Other Authors: School of Electrical and Electronic Engineering
• Format: Theses and Dissertations
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/61628
• Institution: Nanyang Technological University

Priming is an implicit memory effect which has effects on a person’s attitude and evaluation towards an image. Previous study of priming effect involves a lot of self-evaluation questionnaires. In this project, effects of subliminal priming were studied from the perspective of ERP. EEG data was recorded from forty subjects of positive, negative and neutral priming. A series of pre-processing steps including epoch extraction, re-referencing, independent component analysis and artifacts rejection were applied. The study focus on an early response difference which is between 0-100ms and a late ERP component which is between 300-500ms. Quantitative analysis of ERPs was performed. Shift-invariant multi-linear decomposition analysis was used to align ERP data. Comparison between normal averaged ERP and shift CP ERP was made throughout the study. To differentiate the three priming conditions, statistical analysis, feature selection and discriminant analysis using SVM were carried out based on processed ERPs.