Poster Presentation

Poster E in Poster Session E: Thursday, August 6, 10:30 am – 12:15 pm, Kimmel Center, Shorin & Rosenthal Rooms

Temporal Dynamics of Spatial Frequencies in Visual Word, Object, and Place Recognition

Clémence Bertrand Pilon¹, Gabriela Milanova¹, Anjali Singh¹, Martin Arguin¹; ¹Université de Montréal

Presenter: Clémence Bertrand Pilon

It is widely believed that, in visual recognition, low spatial frequencies (SFs) are processed first, then followed by higher SFs. This view is known as the coarse-to-fine theory (Barr, 2003). We studied the temporal dynamics of spatial frequency (SF) processing in visual recognition in three experiments involving word (Exp.1), object (Exp.2) and place (Exp.3) recognition in normal adult observers. The target stimuli were made of an additive combination of the signal (target image) and of a visual white noise patch wherein the signal-to-noise ratio (SNR) varied randomly across stimulus duration. Four SF conditions were defined, with center frequencies of 1.2, 2.4, 4.8 and 9.6 cycles per degree. The results indicate a complex, non-sequential pattern of SF processing which varies across stimulus classes. In Exp.1 (words), the highest SF range dominates early processing, with a shift toward lower SFs later on. In Exp.2 (objects), initial processing was dominated by the 4.8 cpd band, followed briefly by 9.6 cpd, then 1.2 cpd, and finally 2.4 cpd. In Exp.3 (places), processing was dominated first by the 2.4 cpd band, followed by 9.6 cpd and 4.8 cpd bands. These findings challenge the coarse-to-fine theory and show that SF processing order varies across object categories. In each experiment, a machine learning algorithm successfully classified (with >90% accuracy) individual data patterns according to SF condition. This means that SF bands are processed by mechanisms that are dissociable based on their temporal features.

Topic Area: Computational Models of Vision & Visual Cortex