Emotion recognition through cardiovascular response in daily life using KNN classifier

Abstract

Emotion in daily life is difficult to recognize due to disadvantageous of continuous measurement. This study was to develop the method for recognizing daily emotion from a measurement of daily cardiovascular response by using the developed wireless sensor. Seven subjects assessed subjective emotions based on Russell’s emotional circumplex model every 3 h wearing a photo-plethysmography (PPG) sensor. The heart rate variability (HRV) according to two emotional dimensions were tested by the Kruskal-Wallis test. Significant parameters of them were determined to be distinguished among emotions and were applied to recognize emotions using the K-Nearest Neighbor (KNN) algorithm. The arousal and valence were recognized with respective 88.2% and 56.2% accuracy. The methods in this study is extended to monitor and recognized in industrial domain and health care domain requiring recognition of long-term emotion.