Handwriting Based Personality Traits Identification Using Adaptive Boosting and Textural Features

Abstract

Computer analysis of personality traits through handwriting product is becoming increasingly an important thing, mainly because of the deep integration of AI techniques in many fields, such as recruitment services, pedagogy, and mental health diagnostics. Various research studies have shown that there are numerous dimensions of information which can be extracted from a writer's handwriting product. This information had helped reveal writer's gender, identity, age, and also several personality features. Hence, our work aimed at identifying the personality traits of a writer according to Five Factor Model (FFM), by exploiting the textural features of his handwriting product. Thus, the Edge Hinge technique was introduced to extract the textural information found in handwriting images. The exploited dataset for the evaluation of this work consists of a new corpus, dedicated to the experience of the personality traits problem on a group of 285 subjects by instrumentalizing FFM approach. We used several classifiers such as, the Random Forest, Support Vector Machine, and Adaboost in order to choose the best one. The experimental work resulted in higher identification rates than those in the literature, through introducing a combination of classifier technique based on Adaboost to improve performance.