This page is slightly outdated, an updated version will be available soon. In the meantime, please refer to our recent publications for an overview of our current research.
Multistable Perception & Decision-Making
During sustained presentation of an ambiguous stimulus, an
individual's perceptual experience will generally switch between the
different possible alternatives rather than stay fixed on one
interpretation. In a recent study (Einhäuser et
al. 2008) we demonstrated that switches in a variety of such
rivalry stimuli (Necker cube, plaid motion segregation, structure from
motion and auditory stream segregation) are robustly accompanied by
increases in pupil diameter. Since pupil dilation under constant
illumination reflects the activation of the locus coeruleus (LC) - the
brainstem nucleus responsible for synthesis and release of
noradrenaline (NA) throughout the cortex - our data suggest the
involvement of the LC-NA system in rivalry. This is, the LC-NA complex
may play exactly the same role in perception as it is understood to be
playing in behavioral selection. Consequently, we hypothesize that
resolving perceptual ambiguity may be understood as a form of
decision-making. In current research, we are investigating the relation
between such perceptual and behavioral decision-making
processes.
Main Collaborators: Olivia
Carter (Harvard); Christof Koch
(Caltech)
Attention in Natural Scenes
In order to cope with the wealth of information contained in a
natural scene, human observers typically allocate their processing
resources by shifting attention to subsets of the visual stimulus. To
understand the neural processes underlying this attention allocation,
our lab follows three complementary paradigms. First, we measure eye
movements, while observers view modified natural scenes and perform
different tasks (e.g., Einhäuser
& König, 2003; Einhäuser, Rutishauser
& Koch, 2008). Second, we engage observers in rapid recognition
tasks to probe the limits of their recognition performance under
different attentional conditions (e.g.,
Einhäuser, Koch & Makeig, 2007) Finally, we try and transfer
mathematical models of attention to predict rapid scene recognition
(e.g., Einhäuser, Mundhenk et
al., 2007). In our current research, we are aiming at modeling
object recognition and spatial attention in a common framework, to
foster our understanding of the underlying neuronal circuitry and to
improve state-of-the-art computer vision algorithms.
Main Collaborators: Laurent Itti (USC, Bayesian models of
attention); Scott
Makeig (UCSD, EEG in rapid scene recognition); Merielle
Spain/Pietro Perona (Caltech, object recognition); Ueli
Rutishauser/Christof Koch (Caltech, effects
of task), Peter König (University of
Osnabrück, scene statistics)
Eye movements during natural exploration
The oculomotor system is the best-studied motor system in humans and
the direction of gaze serves as an experimentally accessible correlate
for the direction of spatial attention (see above). Although the
direction of gaze under natural conditions results from the combined
movement of body, head and eyes, typical laboratory experiments
restrain the observer and therefore focus on eye-in-head movements. In
contrast, we use a novel, wearable system ("EyeSeeCam") that
continuously aligns a pivotable camera with the direction of the
observer's gaze. The thus obtained gaze-aligned data during free
exploration allows us to reassess models of eye movement control and
eye head coordination for truly free behavior (
Einhäuser, Schumann et al., 2007)
Main Collaborators: Erich Schneider (LMU
Munich); Frank Schumann, Peter
König (University of Osnabrück)
Funded in part by: BaCaTec
Coding Principles in Human and Computational Vision
Can a small number of general coding principles explain many properties of the visual system? In a number of studies we address this issue and link the principle of temporal coherence to complex cell properties in Primary visual cortex (Kayser et al. 2001; Einhäuser et al. 2002; Körding et al 2004), the extraction of complemetary features (Einhäuser et al. 2003), the learning of texture representations and the generation of representations that are well-suited for invariant object classification (Einhäuser, Hipp, et al., 2005). We extended the same principle to somatosensory representations. (Hipp, Einhäuser, et al., 2005). In current research we try to link these results to hierarchical models of object recognition and attention.
C-Code for temporal coherence simulations is available, just send an email
