Palm Print Based Person Verification Using Inception-Resnet-V2 Convolutional Neural Network
Meo Vincent C. Caya, Analyn Niere Yumang, Quinjan Robert R Ocampo, Darien Rhyce B Soria, Rih-Luh Chung, Wen‐Yaw Chung · 2023
Biometric-based person verification system has been a dominant means for verifying a person's uniqueness. Biometrics uses physiological and behavioral characteristics in order to identify a person, such as a fingerprint and iris image. Compared to other physiological characteristics, palm print is used in this study because it is easier to capture, which can be obtained using any low-resolution device, compared to iris or fingerprint images, that need a high-resolution device to capture. The proponents of this study applied the Convolutional Neural Network Inception Resnet V2 in classifying the uniqueness of a person's palm print and studied the effectiveness when applied with different pre-processing techniques. Upon completion of the experiment, it was determined that using 2D Histogram with Morphological closing yields the highest accuracy rate of 87.04% among the three different pre-processing techniques. The researchers also note that a better, more extensive, and more diverse dataset would increase the system's accuracy.