A Comparative Analysis of Evaluating Emotion Recognition Using Arduino-Based Sensors and Consumer Wearable Devices

Abstract

The importance of physiological signals in affective computing research has showcased the use of wearable devices for emotion detection. This study presents a comparative examination of three core emotions: happy, fear, and sad. The two-phase methodology has been used for detection of emotions. At first, from the traditional methodology, non-invasive biomedical sensors along with Arduino UNO R3 were used to record the three physiological signals, such as oxygen saturation, skin conductance, and pulse rate. Phase 2 was developed as a comparative analysis, serve to determine the possibilities for the application of consumer-level smartwatches in obtaining physiological data regarding emotions. Supervised machine learning models such as k-nearest neighbor, support vector machine, and naïve bayes were utilized and gave 98%, 97%, and 96% accuracy, respectively, for phase 1. While phase 2 obtained a much lower percentage due to scattered results.