Conclusions

We have determined that by combining a deep convolutional neural network pre-filter to a minutiae matching fingerprint verification algorithm we have achieved a improved accuracy of approximately 90-95%. Which may not be remarkable to traditional biometric systems, as currently our technology is limited to optical images. But there are many upcoming technology pertaining to fingerprint biometrics such as in the case of smartphones, Qualcomm is coming up with ultrasound sensors which can pass through solid objects to capture information from our fingertips. These technologies may have several noise factors associated to them. Since noise is directly related to high \(FAR\) our incorporation of a pre-filter CNN will reduce it substantially to maintain security to current authentication standards.