Using Biometric Recognition in Residential Security

Abstract

Facial recognition in biometric security is becoming increasingly prevalent daily, with technologies being used to unlock many daily products such as Apple iPhones and computers. However, biometric security in fields such as residential security is a concept that has yet to see widespread use. This paper details the design used to replicate a real-life scenario of artificial intelligence-aided biometrics verification in the context of residential home security. Data is collected via a web app to train a proper model. The app collects biometric data on the user via three videos of three critical angles required in facial recognition models: frontal face, left profile face, and right profile face. A program then samples through a specific number of frames in each video, which is stored as images in an AWS S3 Bucket database to be queried for model training. A Convolutional Neural Network (CNN) framework called DeepFace is utilized to use the data for the facial verification job. The Deepface framework employs a Euclidean distance algorithm to determine the similarity between faces in the database and the face captured for biometric verification, thus determining whether a user is authorized to enter a residence. Once a user is verified, they can enter their house without requiring a key, making access to the residence more convenient.