Signature recognition through spectral analysis

摘要

Features such as shape, motion and pressure, minutiae details and timing, and transformation methods such as Hadamard and Walsh have been used in handwritten signature recognition with various degrees of success. Others have successfully used nonlinear warping functions to optimally time-match an unknown to a standard signature.In this research, a fast Fourier transform is used to transform normalized signatures into the frequency domain. Fifteen harmonics having the largest magnitudes normalized by their corresponding variances were selected and used in a stepwise discriminant analysis. It results in an error rate of 2.5 per cent with the generally more conservative jackknife procedure yielding the same small error rate.