Detection of Handwriting Characteristics using Convolutional Neural Networks

This paper presents an application based on image processing techniques that use deep neural networks in the field of psychology. The approach proposed in the paper automates the well-known methods of determining the main personality traits using graphological analysis (handwriting analysis). The study is based on three characteristics of handwriting: baseline, word slant and writing pressure. To achieve this goal, three VGG-16 Convolutional Neural Networks were trained using a database of handwriting samples. The results obtained are promising, with an accuracy between 78% and 92%. Also, the image processing speed (after training the network) is very high with values around 60 sec.