Automatic Online Signature Verification Based Only on FHE Features: An Oxymoron?

This paper evaluates the feasibility of using only Forensic Handwriting Experts (FHEs) based features for automatic online signature verification. Both, global features and features based on the wavelet representation of the time functions associated with the signing process, which are relevant to FHEs, are considered in this paper. Two combination approaches of global and time function FHE based features are proposed. One of them, consists in a pre-classification of the signatures based on FHE global features so that gross forgeries can be discarded, followed by a Random Forest (RF) classifier using time function based FHE features. The other one, consists in a decision level fusion of two RF classifiers using global and time function FHE based features, respectively. Experimental results on a publicly available database containing Western and Chinese signatures are promising in the sense that automatic online verification systems using exclusively FHE based features achieve verification performances comparable to those of the state-of-the-art over the same datasets.