Tesi etd-04162020-192207
Link copiato negli appunti
Tipo di tesi
Dottorato
Autore
D'ONOFRIO, GRAZIA
URN
etd-04162020-192207
Titolo
Human-Robot Interaction: a Care Robot for Older People with Dementia and Emotion Recognizing by a robotic solution initiative (EMOTIVE) Project
Settore scientifico disciplinare
ING-IND/34
Corso di studi
Istituto di Biorobotica - BIOROBOTICS
Commissione
Membro Prof. CAVALLO, FILIPPO
Presidente Prof. ODDO, CALOGERO MARIA
Presidente Prof. ODDO, CALOGERO MARIA
Parole chiave
- Human-Robot Interaction
Data inizio appello
23/11/2020;
Disponibilità
parziale
Riassunto analitico
Introduction: The present thesis shows two phases of a PhD work during three years and focuses on the interaction of dementia patients with a care robot and the emotion detection analysis. Human-Robot Interaction (HRI) is an area of research dedicated to studying, creating, and testing robotic systems for use by or with humans.
The primary objective of this thesis is to design new control strategies to develop a more natural, safer and more adaptable human-robot interaction.
Materials and Methods: in the first phase, 38 people (M=14; F=24) with AD were screened for eligibility and all were included. The following tests were administered Pre and Post interactions with the robot: Observational Measurement of Engagement (OME), Mini-Mental State Examination (MMSE), Clock Drawing Test (CDT), Frontal Assessment Battery (FAB), Neuropsychiatric Inventory (NPI), Cornell Scale for Depression in Dementia (CSDD), Multidimensional Scale of Perceived Social Support (MSPSS), 14-item Resilience Scale (RS-14), Quality of Life in Alzheimer's Disease (QOL-AD), Caregiver Burden Inventory (CBI), Tinetti Balance Assessment (TBA), and Comprehensive Geriatric Assessment (CGA) was carried out.
In the second phase, 27 hospital employees (M = 12; F = 15) were recruited to run the experiment by a showing of 60 pictures with positive/negative/neutral valence selected in International Affective Picture System (IAPS) database. The experiment was performed with Pepper robot in two modalities: static and active.
About the experimental phase with Pepper in active modality, a concordant mimic was programmed according to the image types (positive, negative and neutral). During the experimentation, the pictures were shown by a web interface with duration of 7 seconds for each slide. For each picture, the participants were called to perform a subjective evaluation about perceived emotional experience (emotive feedback) by Self-Assessment Manikin (SAM). After the participants had used the robotic solution, the Almere Model Questionnaire (AMQ) and System Usability Scale (SUS) were been administered in order to assess the acceptability, usability and functionality of the robotic solution by participants. The analysis is performed on video recordings. The assessment of 3 types of attitude (neutral, positive and negative) is performed through 2 machine learning classification algorithms: k-nearest neighbours (KNN), and random forest (RF).
Results: In the first phase, during Post-robot interactions, significant improvements were observed in RS-14 (p=0.020).
Considering the age of the people, dementia patients with 68-76 years perceived that they had major social support (MSPSS Total: p=0.016) and friends to support them (MSPSS Fri: p=0.014). Indeed, the younger people (55-67 years) were less depressed (CSDD: p=0.033), and more resilient (RS-14: p=0.003). The people aged 77-85 years perceived they had major family support (MSPSS Fam: p=0.018). The participants were gender and education matched without any statistically significant difference.
In the second phase, according to emotion analysis run on videos recorded, it was shown that RF algorithm, all neutral, positive and negative attitudes had an equal and high precision (mean = 0.98) and F-measure (mean = 0.98), with respect to KNN algorithm. The participant groups that had used the robot in static and active modalities did not differ in all considered variables according to KNN and RF algorithms. Moreover, most of participants had confirmed a high level of usability and acceptability of the robot solution.
Conclusion: Care robot may be a useful tool in mitigating depression and loneliness, while enhancing social connectedness, resilience, and overall quality of life for people with dementia. The robot does not constitute a disturbing factor in the elicitation of emotions. The outcome of all work can be integrated into a robotic platform to automatically assess the quality of interaction and to modify its behaviour accordingly.
The primary objective of this thesis is to design new control strategies to develop a more natural, safer and more adaptable human-robot interaction.
Materials and Methods: in the first phase, 38 people (M=14; F=24) with AD were screened for eligibility and all were included. The following tests were administered Pre and Post interactions with the robot: Observational Measurement of Engagement (OME), Mini-Mental State Examination (MMSE), Clock Drawing Test (CDT), Frontal Assessment Battery (FAB), Neuropsychiatric Inventory (NPI), Cornell Scale for Depression in Dementia (CSDD), Multidimensional Scale of Perceived Social Support (MSPSS), 14-item Resilience Scale (RS-14), Quality of Life in Alzheimer's Disease (QOL-AD), Caregiver Burden Inventory (CBI), Tinetti Balance Assessment (TBA), and Comprehensive Geriatric Assessment (CGA) was carried out.
In the second phase, 27 hospital employees (M = 12; F = 15) were recruited to run the experiment by a showing of 60 pictures with positive/negative/neutral valence selected in International Affective Picture System (IAPS) database. The experiment was performed with Pepper robot in two modalities: static and active.
About the experimental phase with Pepper in active modality, a concordant mimic was programmed according to the image types (positive, negative and neutral). During the experimentation, the pictures were shown by a web interface with duration of 7 seconds for each slide. For each picture, the participants were called to perform a subjective evaluation about perceived emotional experience (emotive feedback) by Self-Assessment Manikin (SAM). After the participants had used the robotic solution, the Almere Model Questionnaire (AMQ) and System Usability Scale (SUS) were been administered in order to assess the acceptability, usability and functionality of the robotic solution by participants. The analysis is performed on video recordings. The assessment of 3 types of attitude (neutral, positive and negative) is performed through 2 machine learning classification algorithms: k-nearest neighbours (KNN), and random forest (RF).
Results: In the first phase, during Post-robot interactions, significant improvements were observed in RS-14 (p=0.020).
Considering the age of the people, dementia patients with 68-76 years perceived that they had major social support (MSPSS Total: p=0.016) and friends to support them (MSPSS Fri: p=0.014). Indeed, the younger people (55-67 years) were less depressed (CSDD: p=0.033), and more resilient (RS-14: p=0.003). The people aged 77-85 years perceived they had major family support (MSPSS Fam: p=0.018). The participants were gender and education matched without any statistically significant difference.
In the second phase, according to emotion analysis run on videos recorded, it was shown that RF algorithm, all neutral, positive and negative attitudes had an equal and high precision (mean = 0.98) and F-measure (mean = 0.98), with respect to KNN algorithm. The participant groups that had used the robot in static and active modalities did not differ in all considered variables according to KNN and RF algorithms. Moreover, most of participants had confirmed a high level of usability and acceptability of the robot solution.
Conclusion: Care robot may be a useful tool in mitigating depression and loneliness, while enhancing social connectedness, resilience, and overall quality of life for people with dementia. The robot does not constitute a disturbing factor in the elicitation of emotions. The outcome of all work can be integrated into a robotic platform to automatically assess the quality of interaction and to modify its behaviour accordingly.
File
| Nome file | Dimensione |
|---|---|
Ci sono 1 file riservati su richiesta dell'autore. |
|