Affective computing based personalized meal recommendation

Abstract

Human emotions vary when they crave foods although not all food items will be appealing in all moods. Learning people’s food preferences and making recommendations based on their emotions of that certain time, is very toilsome. This research has been conducted to build a model that will recommend appropriate meals predicated on a person’s current emotion. In this paper, a system has been proposed that will evaluate a person’s emotion through the Electroencephalogram (EEG) signal and analyses the user’s Like, Feelings and Excitement affectivity analysis. To that purpose, we arranged trials to record the EEG signal of 25 people utilizing 14 electrodes, connected directly to their scalp. Here, feature extraction techniques which include Fast Fourier Transform (FFT), Short-time Fourier Transform (STFT), DiscreteWavelet Transform (DWT), Hilbert Huang Transform (HHT), Hjorth Coefficient, Spectral Entropy have been performed on four classification Machine Learning algorithms namely Random Forest Classifier (RF), K-nearest Neighbors (KNN), Support Vector Machine (SVM), XGBoost Classifier (XGB), where all of these anal- yses include both subject-dependent and subject-independent approaches.