Age characterization from online handwriting

Age characterization from handwriting (HW) has important applications as it may allow distinguishing normal HW evolution due to age from abnormal HW change, potentially related to a cognitive decline. We propose, in this work, an original approach for online HW style characterization based on a two-level clustering scheme. The first level allows generating writer independent word clusters according to raw spatial-dynamic HW information. At the second level, the writer words are converted into a Bag of Prototype Words that is augmented by a measure of his/her writing stability across words. For age characterization, we harness the two-level HW style representation using unsupervised and supervised schemes, the former aiming at uncovering HW style categories and their correlation with age and the latter at predicting age groups. Our experiments on a large database show that the two level representation uncovers interesting correlations between age and HW style. The evaluation is based on entropy-based information theoretic measures to quantify the gain on age information from the proposed two-level HW style representation.