Identity Verification through Face Recognition Implemented on Raspberry Pi Framework

Abstract

The influence of identity verification in mobile application have attracted significant attention and growth to mobile technology development. So far, there are various methods involving identity verification in mobile application development, yet, it was still considered challenging and have been a primary study due to its limitation to memory allocation and the computing power.  Subsequently, to address these issues there are different algorithms that were designed for identity verification to adopt the mobile environment and to resolve the arising challenges. In this paper, we proposed a novel,  cost-effective and energy-efficient framework by introducing a mobile-based identity verification on offline-mode using the Raspberry Pi framework. A Raspberry Pi device is wirelessly connected to mobile phone to process the face detection and face verification.  The proposed method is implemented on the latest version of Raspberry Pi 3 model B+ version run in Python 3.7 where the datasets and training sets images were experimented and tested using LBP algorithms for face detection and face verification. With the experimental test result using the confusion matrix in a multiclass classification,  the proposed method showed a results of 87.5% accuracy score,  88% in terms of precision,  88% recall score and 86% F1 score.  In addition, the experimental test were done using 3000 images in a controlled/unconstrained environment, were 20% or 600 of the images were used as data sets. During the offline mode testing, the face detection and verification has resulted to an average timing of 4.98 seconds. Thus, it concludes the feasibility to implement face verification system in the Philippine government services such as the voting system, road check-point, driver’s ID verification to name a few.  Keywords: Raspberry Pi, Local Binary Pattern (LBP), Face Detection, Face Verification, Mobile Application, Identity Verification