Online Signature Verification using Hybrid Features

Biometrics is the process by which a person's physical, behavioral and psychological traits are identified and recorded by an electronic device of identity. Online handwritten signature verification is a biometric based security system. Most of the previous work is based on statistics of online signatures, local and global features. We proposed to test hybrid features on Japanese online dataset from ICDAR2013 [38]. We extract features in frequency domain and time domain as well by using combination of global, Fourier transform and wavelet transform based features. Accuracy is calculated to compare the efficiency of proposed method. It shows that combination of global, DWT and FFT based features yields better results than other combinations. The accuracy achieved by our system is 73.49% which is better than previous systems evaluated on Japanese online dataset.