Non-Gaussian randomness outside and inside the bat’s brain

Abstract

Reflector placement in the habitats of bats is pronouncedly random. Furthermore, the impulsive nature of reflection sequences generated by facets found, e.g., in foliage can endow the probability density function of the received echo amplitudes with heavy tails. Two useful features for classification of foliages, which may be utilized by bats as ubiquitous landmarks, are the extent to which the amplitude distribution deviates from Gaussian (its ‘‘peakiness’’) and an eventual nonstationary gain of the propagation channel. The signal processing in the bat’s auditory system does not have direct access to the target’s reflection sequence, however. Even the shortest echolocation pulses seen in bats are chirps with durations, which are nonnegligible compared to the duration of the reflection sequence and the temporal spacing of salient peaks inside of it. On reception, echoes are subjected to bandpass filtering and demodulation. The latter step prohibits recovery of the reflection sequence by means of pulse compression using a matched filter. Consequently, it is investigated how the demodulated filter bank output may be used directly as a substrate for target classification. Specifically, the possible role of encoding the waveforms in sparse spike trains and comparing binaural information is evaluated.