|
Abstract:
|
The field of "natural scene statistics" is based on the premise that animal nervous systems evolved to interpret the statistics of ethological niches. The statistics of natural scenes are a prior distribution that a sensor (e.g., the eye) samples to acquire information about the world. We used the rat vibrissal (whisker) array, an important model system in sensorimotor neuroscience, to describe tactile scene statistics while incorporating sensor dynamics. Rats use rhythmic (5 - 12 Hz) "whisking" motions of their vibrissae to tactually explore their world. To quantify the statistics of sensory data acquisition through whisking we developed an accurate 3D model of rat whisking dynamics. We then simulated whisking against flat walls and cylinders of different radii while systematically positioning the rat's head to perform an exhaustive search through the full space of possible contact patterns. For each head pose, we quantified all mechanical signals (six forces and moments) at the base of each of the 62 whiskers on the rat's face, and applied a variety of unsupervised learning methods to extract features of the statistical structure that underlies the natural tactile scene.
|
