An Efficient System for Classifying Customer Facial Expressions to Obtain Feedback with Accuracy at Unmanned Restaurants using Improved Support Vector Machine and Flattened Convolutional Neural Networks

The object of the study is to provide an effective feedback system for robotic restaurants which uses customer's facial expressions to acquire ratings about customers experience. Materials and methods: In this study we used the FER2013 dataset inorder to implement the system. The proposed system is built based on an Improved support vector machine algorithm with a novel kernel trick method. This system uses the customer's facial expression in order to obtain the rating. Two sample groups have been considered for study in this article such as Improved support vector machine and Flattened convolutional neural networks with sample size 51 for each. Results: The proposed system performed better by obtaining higher accuracy i.e. 82.14% than the Flattened convolutional neural networks where the accuracy was 61.16% The statistical significance of two algorithms observed is
$\mathbf{p} < 0.001 \ (2- \text{tailed})$
from the independent sample
$\mathbf{T}$
test. Conclusion: From the results it can be clearly seen that Improved support vector machine show higher accuracy than Flattened convolution neural networks.