Biosonar in a textured world

Abstract

Capabilities of bats to analyze spectral profiles have been demonstrated by several researchers. However, suggestions as to the informational content of the biologically accessible features in terms of the echo-generating process are not yet entirely convincing. Invoking a general analogy to visual color perception does not provide an operational definition of the hypothetical perceptual mechanism. The proposed mechanisms are mostly restricted to a scenario, where the spacing of two reflectors may be determined via the resulting comb-filter transfer function. The generality of this concept is curtailed by most natural surfaces constituting aggregates of multiple reflecting facets with a pronounced random component to their arrangement. Classification of natural textures is conceived to be a pivotal capability which may underly performance in several tasks probably performed by the animals, e.g., landmark identification and control of scanning movement. In order to understand how the demands of texture classification relate to the key features of biological sonar systems, the following steps are taken: Starting from a database of textures typical of bat’s habitats (i.e., foliage), the implications of two paramount attributes of biological sonar systems are studied: sensor mobility and a joint time-frequency representation encoded in spike patterns.