Automatic pain assessment can be defined as the set of computer-aided technologies allowing to recognise pain status. Reliable and valid methods for pain assessment are of primary importance for the objective and continuous monitoring of pain in people who are unable to communicate verbally. In the present work, we propose a novel approach for the recognition of pain from the analysis of facial expression. More specifically, we evaluate the effectiveness of Graph Neural Network (GNN) architectures exploiting the inherent graph structure of a set of fiducial points automatically tracked on subject faces. Experiments carried over on the publicly available dataset BioVid, show how the proposed method reaches higher levels of accuracy when compared with baseline models on acted pain, while outmatching state of the art approaches on spontaneous pain.
Deep graph neural network for video-based facial pain expression assessment / S. Patania, G. Boccignone, S. Bursic, A. D'Amelio, R. Lanzarotti - In: SAC '22: Proceedings / [a cura di] J. Hong, M. Bures, J. Won Park, T. Cerny. - [s.l] : ACM, 2022. - ISBN 9781450387132. - pp. 585-591 (( Intervento presentato al 37. convegno Symposium on Applied Computing tenutosi a Brno nel 2022 [10.1145/3477314.3507094].
Deep graph neural network for video-based facial pain expression assessment
S. Patania
;G. Boccignone;S. Bursic;A. D'Amelio;R. Lanzarotti
2022
Abstract
Automatic pain assessment can be defined as the set of computer-aided technologies allowing to recognise pain status. Reliable and valid methods for pain assessment are of primary importance for the objective and continuous monitoring of pain in people who are unable to communicate verbally. In the present work, we propose a novel approach for the recognition of pain from the analysis of facial expression. More specifically, we evaluate the effectiveness of Graph Neural Network (GNN) architectures exploiting the inherent graph structure of a set of fiducial points automatically tracked on subject faces. Experiments carried over on the publicly available dataset BioVid, show how the proposed method reaches higher levels of accuracy when compared with baseline models on acted pain, while outmatching state of the art approaches on spontaneous pain.File | Dimensione | Formato | |
---|---|---|---|
ACM_SigConf_SAC2022.pdf
accesso aperto
Tipologia:
Post-print, accepted manuscript ecc. (versione accettata dall'editore)
Dimensione
4.89 MB
Formato
Adobe PDF
|
4.89 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.