TY - GEN
T1 - Dynamic probabilistic CCA for analysis of affective behaviour
AU - Nicolaou, Mihalis A.
AU - Pavlovic, Vladimir
AU - Pantic, Maja
N1 - eemcs-eprint-22966
PY - 2012/10/7
Y1 - 2012/10/7
N2 - Fusing multiple continuous expert annotations is a crucial problem in machine learning and computer vision, particularly when dealing with uncertain and subjective tasks related to affective behaviour. Inspired by the concept of inferring shared and individual latent spaces in probabilistic CCA (PCCA), we firstly propose a novel, generative model which discovers temporal dependencies on the shared/individual spaces (DPCCA). In order to accommodate for temporal lags which are prominent amongst continuous annotations, we further introduce a latent warping process. We show that the resulting model (DPCTW) (i) can be used as a unifying framework for solving the problems of temporal alignment and fusion of multiple annotations in time, and (ii) that by incorporating dynamics, modelling annotation/sequence specific biases, noise estimation and time warping, DPCTW outperforms state-of-the-art methods for both the aggregation of multiple, yet imperfect expert annotations as well as the alignment of affective behavior.
AB - Fusing multiple continuous expert annotations is a crucial problem in machine learning and computer vision, particularly when dealing with uncertain and subjective tasks related to affective behaviour. Inspired by the concept of inferring shared and individual latent spaces in probabilistic CCA (PCCA), we firstly propose a novel, generative model which discovers temporal dependencies on the shared/individual spaces (DPCCA). In order to accommodate for temporal lags which are prominent amongst continuous annotations, we further introduce a latent warping process. We show that the resulting model (DPCTW) (i) can be used as a unifying framework for solving the problems of temporal alignment and fusion of multiple annotations in time, and (ii) that by incorporating dynamics, modelling annotation/sequence specific biases, noise estimation and time warping, DPCTW outperforms state-of-the-art methods for both the aggregation of multiple, yet imperfect expert annotations as well as the alignment of affective behavior.
KW - METIS-296256
KW - IR-84313
KW - EC Grant Agreement nr.: ERC-2007-STG-203143 (MAHNOB)
KW - EC Grant Agreement nr.: FP7/288235
KW - EC Grant Agreement nr.: FP7/2007-2013
KW - EWI-22966
KW - HMI-MI: MULTIMODAL INTERACTIONS
U2 - 10.1007/978-3-642-33786-4_8
DO - 10.1007/978-3-642-33786-4_8
M3 - Conference contribution
SN - 978-3-642-33785-7
T3 - Lecture Notes in Computer Science
SP - 98
EP - 111
BT - Proceedings of the European Conference on Computer Vision, ECCV 2012
PB - Springer
CY - Berlin
T2 - European Conference on Computer Vision, ECCV 2012
Y2 - 7 October 2012 through 13 October 2012
ER -