Applying psychophysical reverse correlation to high dimensional natural stimuli E Joosten, M A Giese |
|---|
Computational Sensomotorics, HIH,CIN,BCCN, University Clinic Tuebingen, Germany
|
| Applying psychophysical reverse correlation to high dimensional natural stimuli. Joosten ERM and Giese MA Psychophysical reverse correlation involves a trial-by-trial analysis of the relationship between stimulus noise and the observers' response. The results (classification images [1]) provide direct insight into the observers' perceptual templates. However, it remains unclear whether noisy images are similarly processed as natural images. METHODS. We developed an extension of psychophysical reverse correlation for high dimensional natural stimuli. With faces from the Cohn-Kanade database [2] stimuli were modeled by active appearance models [3] trained separately on different spatial components. This allowed us to parameterize shape variations (e.g. of the eye or mouth) by low dimensional vectors. With this classifier based approach we determined perceptual templates of different facial expressions. RESULTS AND CONCLUSION. We compared perceptual templates derived from this new approach with classical pixel-based classification images [4]. [1] Ahumada, JVis, 2:1,2002. [2] Kanade et al., Proceedings of the Fourth IEEE International Conference on Automatic Face and Gesture Recognition(FG'00),2000. [3] Cootes et al., IEEE Transactions on Pattern Analysis and Machine Intelligence,1998. [4] Sekuler et al., Curr.Biol., 14(3):5,2004. ACKNOWLEDGEMENTS: This research is supported by the European Commission, Fp7-PEOPLE-2011-ITN(MarieCurie):ABCPITN-GA-011-290011, FP7-249858-TP3 TANGO and FP7-ICT-248311 AMARSi |
| Up Home |
|---|