A Smart Home Security System In Low Computing IoT Environment

This work targets design of a face authenticated door-lock system in Raspberry pi and proposes a different neural network model which directly classifies a person's face without considering the euclidean distance. It is to overcome the limitation of handling a large number of stored faces and the cases of variation in angle and illumination in the face image. A set of fully connected layers on the stored face embeddings is trained to predict a person in the last layer. The proposed neural network model, local binary pattern histogram fully connected face authentication (LBPHFCFA), can memorize and generalize to eradicate the flaws of state-of-the-art algorithms. It also tends to work precisely along with the pipeline in Raspberry Pi without much computation overhead as well as better foils the impersonation attack on the system.