Haar wavelet transform based histogram concatenation model for finger print spoofing detection
Susan Shaju, Diana Davis · 2017
Biometrics in general is basically a measure to analyze the identity of a person by means of the person's bodily characteristics. Biometric authentication has now a day's been a very important part of the digital world. Among the several biometric authentications fingerprints are the most widely accepted one due to its uniqueness factor. But overtime the biometric system has become intensively vulnerable to the various kinds of attacks by the imposters. The finger prints can be easily fabricated using the materials such as gelatin, latex or silicon etc. So here we are proposing a software based design model for fingerprint spoofing detection where the texture features is extracted in wavelet domain. The fingerprint images are partitioned into nxn patches and are then transformed to wavelet domain using the Haar wavelet transforms. The binary pattern generated by comparison of the approximation sub band coefficients is then transformed into histograms. The concatenation of the histograms results in the corresponding feature vector that represents the textural characteristics. And finally based on these histogram features the test fingerprints are classified to be either spoofed or real using Support vector machine.