FH-SSTNet: Forehead Creases based User Verification using Spatio-Spatial Temporal Network

TL;DR

A new CNN model called the Forehead Spatio-Spatial Temporal Network (FH-SSTNet), which utilizes a 3D CNN architecture with triplet loss to capture distinguishing features and enhances the model's discrimination capability using Arcloss in the network's head.

Abstract

Biometric authentication, which utilizes contactless features, such as forehead patterns, has become increasingly important for identity verification and access management. The proposed method is based on learning a 3D spatio-spatial temporal convolution to create detailed pictures of forehead patterns. We introduce a new CNN model called the Forehead Spatio-Spatial Temporal Network (FH-SSTNet), which utilizes a 3D CNN architecture with triplet loss to capture distinguishing features. We enhance the model's discrimination capability using Arcloss in the network's head. Experimentation on the Forehead Creases version 1 (FH-V1) dataset, containing 247 unique subjects, demonstrates the superior performance of FH-SSTNet compared to existing methods and pre-trained CNNs like ResNet50, especially for forehead-based user verification. The results demonstrate the superior performance of FH-SSTNet for forehead-based user verification, confirming its effectiveness in identity authentication.

Figures (5)

• Figure 1: Forehead creases are characterized by the presence of vertical, horizontal, and diagonal lines under facial expression, which serve as boundary markers along with the textured patterns observed on the skin's surface. In human forehead anatomy, forehead creases result from their connection to the muscles below the eyelids, Frontal muscles create facial expressions, like raising eyebrows and wrinkling the forehead, Frontal-is muscles meet in the middle, causing forehead wrinkles and Age, Sun exposure, and facial expression cause noticeable lines.
• Figure 2: A pre-processing procedure to illustrate the utilization of images in the time domain for learning spatio-spatial temporal features. In the first step, we localize and segment the forehead area and then divide it into small-sized overlapped patches with a stride of 5. In the second step, all patches are stacked in the third dimension sequentially to transform them into spatiotemporal format.
• Figure 3: Block diagram of the proposed FH-SSTNet for forehead -based person verification. In the first step, a stacked forehead patched into video resulting in 3D representation called a non-local spatio-spatial-temporal relationship is fed into the Backbone network, which generates the non local spatio-spatial temporal features. The following trainable fully connected layer (Head), allows them to process and more discriminate the non-local spatio-spatial temporal feature using Arcloss.
• Figure 4: Examples of forehead images corresponding to different subjects from FH-V1 dataset.
• Figure 5: DET curve for comparative analysis among four different model. X-axis indicates the false match rate and y-axis indicates the false non match rate of forehead creases (FH-V1) dataset.