Publicação
Heart rate estimation using video in psychology experiments
| Resumo: | Heart rate is a relevant physiological marker used in several areas, namely psychology, as a measure of the anxiety and stress among other states. Typically, the heart rate is calculated from ECG that implies using dedicated equipment with electrodes placed on the human subject which can be considered invasive in many situations i.e. not comfortable or humanly suitable. With the advances in computer vision several works proposed methods to estimate the heart rate from video capturing skin patches of the subjects (e.g. for head, overall face, …). However, although promising results there no conclusive proofs on the accuracy and applicability in more realistic conditions (e.g. outside of the laboratory) namely due to the very controlled scenarios or limited sampling time. In this dissertation we proposed to evaluate the usefulness of heart rate estimation based on video and built upon the state of the art to address more realistic and challenging conditions i.e. less controlled scenarios and evaluate it under larger monitoring sessions (>1 minute). We performed two experiments based on video stimulus where the objective was to measure the HR changes induced by the video. In both scenarios, ECG was used to extract the HR that was used as ground truth. The first scenario was acquired with videos to elicit disgust (25 minutes), the second using smaller videos (<1 minute) using a neutral and “happiness” inducing videos. Our results show that the heart rate estimation is very sensitive to noise and not clear relation on the complete studies was observed in any of the scenarios. However, when studying the relation between the HR estimated from video and from ECG it was clear that both were highly correlated in limited time intervals suggesting that video estimated HR may be worthy to explore. In the process we developed PsyVidLab that besides incorporating the video estimated HR allows synchronous acquisition of video, ECG and some basic image processing modules namely emotion estimation from facial expression. |
|---|---|
| Autores principais: | Assunção, Luís Pedro Santos de |
| Assunto: | Ritmo cardíaco - Monitorização Electrocardiografia Visão por computador Redes neuronais Psicologia |
| Ano: | 2016 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Aveiro |
| Idioma: | inglês |
| Origem: | RIA - Repositório Institucional da Universidade de Aveiro |
| Resumo: | Heart rate is a relevant physiological marker used in several areas, namely psychology, as a measure of the anxiety and stress among other states. Typically, the heart rate is calculated from ECG that implies using dedicated equipment with electrodes placed on the human subject which can be considered invasive in many situations i.e. not comfortable or humanly suitable. With the advances in computer vision several works proposed methods to estimate the heart rate from video capturing skin patches of the subjects (e.g. for head, overall face, …). However, although promising results there no conclusive proofs on the accuracy and applicability in more realistic conditions (e.g. outside of the laboratory) namely due to the very controlled scenarios or limited sampling time. In this dissertation we proposed to evaluate the usefulness of heart rate estimation based on video and built upon the state of the art to address more realistic and challenging conditions i.e. less controlled scenarios and evaluate it under larger monitoring sessions (>1 minute). We performed two experiments based on video stimulus where the objective was to measure the HR changes induced by the video. In both scenarios, ECG was used to extract the HR that was used as ground truth. The first scenario was acquired with videos to elicit disgust (25 minutes), the second using smaller videos (<1 minute) using a neutral and “happiness” inducing videos. Our results show that the heart rate estimation is very sensitive to noise and not clear relation on the complete studies was observed in any of the scenarios. However, when studying the relation between the HR estimated from video and from ECG it was clear that both were highly correlated in limited time intervals suggesting that video estimated HR may be worthy to explore. In the process we developed PsyVidLab that besides incorporating the video estimated HR allows synchronous acquisition of video, ECG and some basic image processing modules namely emotion estimation from facial expression. |
|---|