Handwritten Signature Verification Using Hidden Markov Model with Hybrid Wavelet Transform

Abstract

Handwritten online signature offers x, y, z coordinates along with pressure, azimuth and altitude of the pen tip. Hybrid wavelet transform was applied on the first 128 samples of the pressure parameter and 1- 16 and 33- 64 samples of the output were used as feature vector for signature verification. Using Hidden Markov Model (HMM) classifier with ergodic model, the performance was compared. For 1-16 samples DHT KEKRE offers best performance of FRR 1 % and FAR 1 %. For 33-64 samples, KEKRE 128 and HADAMARD HAAR offers best performance with FRR 1 % and FAR 1 %. Feature vector with 33-64 samples offers better FRR-FAR than 1–16 samples. For 1–16 samples DCT HAAR offers best performance of AAR 9% and ARR 3%. For 33-64 samples, DHT 128 offers best performance with AAR 34% and ARR 23%. 33-64 samples offers better AAR - ARR than 1–16 samples.