390 likes | 474 Views
E. Gilet (1) , J. Diard (2) , R. Palluel-Germain (2) , P. Bessière (1) (1) Laboratoire d’Informatique de Grenoble – CNRS, France (2) Laboratoire de Psychologie et NeuroCognition – CNRS, France July, 5, 2010 http://diard.wordpress.com/ Julien.Diard@upmf-grenoble.fr.
E N D
E. Gilet(1), J. Diard(2), R. Palluel-Germain(2), P. Bessière(1) (1) Laboratoire d’Informatique de Grenoble – CNRS, France (2) Laboratoire de Psychologie et NeuroCognition – CNRS, France July, 5, 2010 http://diard.wordpress.com/Julien.Diard@upmf-grenoble.fr BayesianAction-Perceptionloopmodeling: Application to trajectorygeneration and recognition usinginternalmotor simulation
Perception of actions (Calvo-Merino et al., 2004)
Reading and writing letters Writing Reading letters Reading pseudo letters (Longcamp, 2003)
Interpretation • Motor simulation of actions during perception • Articulation between perception and action processes
Modelingbothreading and writingModelinginternal simulation of movements
Summary • BAP model • architecture and definition: overview • Experimentalresults • simulation of cognitive tasks • Experimentalprediction
BAP model structure internalletterrepresentation coherence variables perception model simulated perception model action model
Cartesian and effector spaces • Common space for perceptive and motor internal representations • Cartesianspace
Letterrepresentation « Laplace succession laws »
Perception model • Deterministicvia-point extraction
Trajectorygeneration model • Minimum-acceleration model: • Costfunction • Boundary conditions • Polynomial solution
Simulated perception model • Identical to the perception model
Coherence variables • Allow to activate or deactivatesubmodels • « Bayesianswitch »
Coherence variable for controllingsubmodel activation • Model • λbinary variable • Joint • Inference • P(A) = P(A): value of Bdoes not influence A λ A B A B A B
Summary • BAP model • architecture and definition: overview • Experimentalresults • simulation of cognitive tasks • Experimentalprediction
Perception: reading letters Correct recognition: 93.36%
Perception: writer recognition Correct recognition: 79.5%
Action: writing letters Variabilitybetween trials Variabilitybetweenwriters
Motor equivalence • Writer “style” • (Wright, 1990) • Common activated motor areas • (Wing, 2000) (Serratrice. 1993)
Perception and Action: Copy Trajectory copy Letter copy
Perception and Action: Reading letters with motor simulation Recall: readingletterswithout simulation
Perception and Action: Reading letters with motor simulation
Perception and Action: Reading letters with motor simulation • Complete trajectories • Correct recognition score with simulation 93.36% • Correct recognition score without simulation 90.2% • Incompletetrajectories
Summary • BAP model • architecture and definition: overview • Experimentalresults • simulation of cognitive tasks • Experimentalprediction
Preliminary data F(1,23) = 3.06, p = 0.093
Summary • BAP model • Bayesian model of perception and action • Includes an internal simulation loop • Cognitive tasks • Reading without and with motor simulation • Writer recognition • Writing with different effectors • Copying letters and trajectories • Basis for experimental predictions
Thank you for your attention ! Questions ?