Real-time Evaluation of Food Acceptance From Facial Expressions Based on Exponential Decay

This research paper proposes a novel method for estimating food acceptance from real-time food consumption video streaming under partially occluded facial expressions. The facial expressions are evaluated based on the Facial Expression Recognition (FER) system. Here, the occlusion is identified as a spoon, fork, or hand. And the Object Detection API from TensorFlow is used as an occlusion detection tool. The combination of the Exponential Decay and the Bayes’ theorem (called EB algorithm) is used to estimate the probabilities of food acceptance from the occluded facial expressions. The EB algorithm also calculates the facial reactions towards food tastes such as bitterness, sourness, sweetness, umami, and saltiness to predict the likelihood of food acceptance. The simulations and the customer comparison results indicate that the presented food acceptance system is one of the accurate ways to signify food acceptance in a real-time food environment.