Online handwritten signature verification based on association of curvature and torsion feature with Hausdorff distance

DeepFuseOSV: online signature verification using hybrid feature fusion and depthwise separable convolution neural network architecture

“…Realising one‐shot learning by the proposed OSV framework qualifies it to deploy in real time scenarios in which gaining more signature samples is impossible, for example, such as e‐commerce and m‐commerce applications and so forth. Similarly, the tables confirm that only a few works [2, 23] evaluated their proposed frameworks with S_10, S_15, R_10, R_15 categories in case of MCYT, R_10, R_15 categories of SVC and R_05 and R_10 categories of SUSIG datasets.…”

“…On selecting the optimal set of parameters, to appraise the proposed OSV framework, we have performed a wide set of experiments which covers every single possible category of experimentation using three broadly utilised publicly accessible datasets, i.e. MCYT‐100 [1, 2, 4, 5, 35], SVC – Task 2 [24, 36, 37], SUSIG [3, 26, 27]. Recently Ruben et al [20] developed a novel largest online signature dataset named ‘DeepSignDB’, which is a collection of signature samples from widely used datasets like MCYT‐330, BiosecurID, Biosecure DS‐2, e‐BioSignDS1 and e‐BioSignDS2.…”

“…As exhibited in Tables 5–7, regardless of the systems proposed in [2, 5, 9, 11, 12, 25, 35, 43] is yielding better EER values contrasted with the proposed framework, the essential pitfall with these models is that these models are assessed with only one category, i.e. S_01, R_01.…”

“…Whereas, the proposed model is thoroughly evaluated with wide experimentation in Skilled 1, 5, 10, 15, 20 and Random 1, 5, 10, 15, 20. Hence, its superiority is validated as compared to [2, 5, 9, 11, 12, 25, 35, 43].…”

“…are prone to risks like stealing or forgetting them. Hence, online signature verification (OSV) is widely used in financial, e‐commerce and m‐commerce applications to legally bind an individual by detecting the forgery of online signatures [1–3]. As presented in Fig.…”