Study on the usage feasibility of continuous-wave radar for emotion recognition. (April 2020)

• Record Type: Journal Article
• Authors: Gouveia, Carolina; Tomé, Ana; Barros, Filipa; Soares, Sandra C.; Vieira, José; Pinho, Pedro
• Is Part Of: Biomedical signal processing and control. Volume 58(2020)
• Publication Date: 2020-04
• Subjects: Continuous wave radar -- Emotion recognition -- Pattern recognition -- Support-vector machine -- K-nearest neighbour -- Random Forest; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101835
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Emotion recognition through the respiratory signal, acquired by radar-based system. Three classification algorithms were compared: SVM, KNN and Random Forest. Emotion identification was achieved with accuracy rates between 60% and 70%. Beside the vital signs analysis, also with other types of body motions were considered. Abstract: Non-contact vital signs monitoring has a wide range of applications, such as in safe drive and in health care. In mental health care, the use of non-invasive signs holds a great potential, as it would likely enhance the patient's adherence to the use of objective measures to assess their emotional experiences, hence allowing for more individualized and efficient diagnoses and treatment. In order to evaluate the possibility of emotion recognition using a non-contact system for vital signs monitoring, we herein present a continuous wave radar based on the respiratory signal acquisition. An experimental set up was designed to acquire the respiratory signal while participants were watching videos that elicited different emotions (fear, happiness and a neutral condition). Signal was registered using a radar-based system and a standard certified equipment. The experiment was conducted to validate the system at two levels: the signal acquisition and the emotion recognition levels. Vital sign was analysed and the three emotions were identified using different classification algorithms. Furthermore, the classifier performance was compared, having in mind the signal acquired by both systems. Three different classification algorithms were used: the support-vector machine, K-nearest neighbour and the Random Forest. The achieved accuracy rates, for the three-emotion classification, were within 60% and 70%, which indicates that it is indeed possible to evaluate the emotional state of an individual using vital signs detected remotely.