SignatureFeatures: A Dynamic Signature Feature Dataset Derived from ATVS-SSig DB

SignatureFeatures is a derived dynamic signature feature dataset created from the ATVS-SyntheticSignature Database (ATVS-SSig DB). The original ATVS-SSig DB contains online synthetic signature trajectories stored with X-coordinate, Y-coordinate, timestamp, pen-up indicator, and pressure information. In this dataset, the original trajectory records were cleaned, processed, and transformed into a machine-learning-ready feature dataset. The final SignatureFeatures dataset contains 11,466 usable signatures from 215 users, comprising 6,986 genuine samples and 4,480 forgery samples generated through a cross-user random forgery assignment protocol. Each signature trajectory is represented using ten dynamic features: X, Y, vx, vy, ax, ay, dx, dy, theta, and distance. The dataset is suitable for dynamic signature verification, writer-independent partitioning, feature extraction, deep learning model training, and ensemble-based verification experiments. This dataset is also publicly available on Kaggle:https://www.kaggle.com/datasets/atiyakazi160184/signaturefeatures