Electrodermal activity (EDA) is a property of the human body affected by changes in the skin resistance due to the effects of the sympathetic nervous system (SNS) on sweat glands permeability. Changes in EDA reflect the physiological processes related to emotions, that occur out of human control. In the last decades this biometric signal is gaining interest in the field of Emotion Recognition and for a variety of healthcare applications. One of the main objectives of the research carried out in this field is the development of classification algorithms able to recognize different emotions and to understand which is the kind of stimuli that evoked such response. However, the intra-individual variability of the EDA signal among different subjects makes the achievement of this goal still challenging. To increase the accuracy of classification systems for emotion recognition based on EDA signals, the scientific research is analysing which are the features with the highest information content for this particular signal. In this Thesis, 16 features have been extracted, 8 in the time domain and 8 in the frequency domain, and analysed in order to evaluate which of them has the highest significance and within what terms. The Empatica E4 wearable sensor was used for data acquisition in concomitance with a prototype static sensor, the Arduino UNO based GSR-Grove sensor to obtain simultaneous recordings and to evaluate the E4 sensor against the GSR-Grove one. The experiment focused on the ability of these sensors to predict relative physical activity intensity. The 4 enrolled subjects were asked to perform activities with 3 different levels of intensity: resting, medium and high intensity exercise. The obtained dataset consisted of 36 recordings for each sensor. Data were processed and analysed in MATLAB. What emerges from the results is that features’ significance cannot be generalised for all the subjects and it is strictly dependent on intra-individual characteristics. There are features that result to be non-significant at all, but this should be confirmed by involving a larger population. Enlarging the study to a greater number of subjects could be promising in order to identify common EDA patterns for subclasses of subjects. This way it would be possible to design subject-dependent classification algorithms. Focusing on the two involved sensors, the Empatica E4 sensor resulted to be more reliable. This is an encouraging outcome for the future of EDA processing within the field of wearables.

Experimental Evaluation of a Wrist-Worn Device for the Measurement of the Galvanic Skin Response

PIERSANTI, AGNESE
2018/2019

Abstract

Electrodermal activity (EDA) is a property of the human body affected by changes in the skin resistance due to the effects of the sympathetic nervous system (SNS) on sweat glands permeability. Changes in EDA reflect the physiological processes related to emotions, that occur out of human control. In the last decades this biometric signal is gaining interest in the field of Emotion Recognition and for a variety of healthcare applications. One of the main objectives of the research carried out in this field is the development of classification algorithms able to recognize different emotions and to understand which is the kind of stimuli that evoked such response. However, the intra-individual variability of the EDA signal among different subjects makes the achievement of this goal still challenging. To increase the accuracy of classification systems for emotion recognition based on EDA signals, the scientific research is analysing which are the features with the highest information content for this particular signal. In this Thesis, 16 features have been extracted, 8 in the time domain and 8 in the frequency domain, and analysed in order to evaluate which of them has the highest significance and within what terms. The Empatica E4 wearable sensor was used for data acquisition in concomitance with a prototype static sensor, the Arduino UNO based GSR-Grove sensor to obtain simultaneous recordings and to evaluate the E4 sensor against the GSR-Grove one. The experiment focused on the ability of these sensors to predict relative physical activity intensity. The 4 enrolled subjects were asked to perform activities with 3 different levels of intensity: resting, medium and high intensity exercise. The obtained dataset consisted of 36 recordings for each sensor. Data were processed and analysed in MATLAB. What emerges from the results is that features’ significance cannot be generalised for all the subjects and it is strictly dependent on intra-individual characteristics. There are features that result to be non-significant at all, but this should be confirmed by involving a larger population. Enlarging the study to a greater number of subjects could be promising in order to identify common EDA patterns for subclasses of subjects. This way it would be possible to design subject-dependent classification algorithms. Focusing on the two involved sensors, the Empatica E4 sensor resulted to be more reliable. This is an encouraging outcome for the future of EDA processing within the field of wearables.
2018
2020-03-13
Experimental Evaluation of a Wrist-Worn Device for the Measurement of the Galvanic Skin Response
File in questo prodotto:
File Dimensione Formato  
Tesi_Piersanti_Agnese.pdf

Open Access dal 13/03/2022

Dimensione 3.77 MB
Formato Adobe PDF
3.77 MB Adobe PDF Visualizza/Apri

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12075/7229