Abstract:Finger vein recognition (FVR) is a technique for identity authentication based on finger vein images (FVIs) that are acquired using a specific device, which has become one of hot spots in the field of biometrics. The main idea of traditional FVR schemes is to directly extract features from FVIs or finger vein patterns (FVPs) and then compare features among FVIs to find the best match. However, the features extracted from FVIs contain much redundant data, while the features extracted from FVPs are greatly influenced by image segmentation methods. Recently, in order to improve the recognition rate and release the high complexity of image preprocessing, a finger vein recognition method based on deep belief network (DBN) with the features extracted from curvature gray images (CGI) has been proposed. However, the training process of DBN is somewhat time-consuming, and the background information of CGI may affect the recognition rate. In order to further improve the accuracy and speed up the training process, a new FVR algorithm based on the improved convolutional neural network (CNN) and curvature gray feature decomposition (CCFD) is proposed in this paper. First, we calculate the curvature of an FVI using a two-dimensional Gaussian template. Then we extract two gray images from the FVI with different scales and add these two images to obtain a CGI. Unlike the previous method, we further decompose this image into two components named vein curvature gray feature image (VCCFI) and background curvature gray feature image (BCGFI). Finally, using VCCFIs as input, an improved CNN is trained and used to recognize the identity of the input FVI. Experimental results show that our scheme is effective and better than traditional schemes and the previous DBN based method.

DeepFinger: A Cascade Convolutional Neuron Network Approach to Finger Key Point Detection in Egocentric Vision with Mobile Camera

Fingertip in the Eye: A cascaded CNN pipeline for the real-time fingertip detection in egocentric videos

Deep Learning Based Hand Gesture Recognition in Complex Scenes

Two-stream Convolutional Neural Network for Accurate RGB-D Fingertip Detection Using Depth and Edge Information.

Attention Based Dual Branches Fingertip Detection Network and Virtual Key System

Curvature gray feature decomposition based finger vein recognition with an improved convolutional neural network

Hand Segmentation and Fingertip Tracking from Depth Camera Images Using Deep Convolutional Neural Network and Multi-task SegNet

Two-Stream Binocular Network: Accurate Near Field Finger Detection Based On Binocular Images

Real-time Fingertip Tracking and Gesture Recognition Using RGB-D Camera

Unified learning approach for egocentric hand gesture recognition and fingertip detection

Accurate Fingertip Detection from Binocular Mask Images.

FingerNet: Pushing The Limits of Fingerprint Recognition Using Convolutional Neural Network

Real Time Fingertip Detection with Kinect Depth Image Sequences.

Anchors Based Method for Fingertips Position Estimation from a Monocular RGB Image using Deep Neural Network

Fv-Net: Learning A Finger-Vein Feature Representation Based On A Cnn

Real-time Fingertip Detection Based on Depth Data

Convolutional Auto-Encoder Based Deep Feature Learning for Finger-Vein Verification

Cascaded hierarchical CNN for 2D hand pose estimation from a single color image

FKIMNet: A Finger Dorsal Image Matching Network Comparing Component (Major, Minor and Nail) Matching with Holistic (Finger Dorsal) Matching

Faster R-CNN with Inception V2 for Fingertip Detection in Homogenous Background Image

Deep Multi-Task Learning based Fingertip Detection