Performance Analysis of CNN Algorithm in Comparison with LR algorithm for Face Recognition in Smart-Lock

The purpose of the proposed research is to compare the accuracy of convolutional neural networks and logistic regression for facial recognition in smart locks. CNN and LR classifiers are applied on the olivetti faces dataset. 40 samples are taken in total and 20 per each group in order to compare the performance analysis of both the algorithms. The values of alpha, beta, percentage confidence and preset power are 0.05, 0.5, 95 and 80 respectively, the enrollment ratio is 1 (Clincalc.com). For the dataset used on the machine learning algorithms, novel CNN produces 94.53% and the logistic regression algorithm produces the accuracy of 92.86% in recognising faces in smart locks. The independent t test significance value obtained is 0.07 which is not less than 0.05 and hence both the algorithms are not statistically significant. The research proves that the algorithm Convolutional Neural Network produces better accuracy than Logistic Regression in recognising face.