Finger knuckle pattern person authentication system based on monogenic and LPQ features

Contemporary biometrics is a swiftly broad field of research. Nowadays, biometrics is predominantly being deployed as a personal identification system in diver's real-world applications. Despite remarkable progress, their performance remains insufficient for security applications. To date, Finger Knuckle Print (FKP) has been explored as a potential biometric characteristic to attain acceptable accuracy and security. This paper presents a novel and efficient scheme to extract features from FKP images, namely Monogenic Local Phase Quantization (M-LPQ), for FKP recognition. First, the monogenic filters are applied to decompose the ROI of FKP images into three complementary parts: the band pass, vertical and horizontal band pass components. Next, we compute the local energy, phase, and local orientation. At that point, LPQ descriptor is endeavoring to encode these complementary parts to compute histograms. These histograms' sequences are concatenated together in the subsequent stage to build an enormous feature vector. To reduce the dimension of the M-LQP features vectors for FKP recognition, Principal Component Analysis and Linear Discriminant Analysis are employed. Ultimately, the Mahalanobis Cosine Distance is used to determine the person's identity. Exploratory outputs show that the introduced framework strikingly achieved lower error rates and yield played out the cutting edge strategies. As a consequence, we were able to get good outcomes by fusing all combinations of four fingers with 99.90 percent Recognition Rate and 0.01 percent EER value.