You are here

Bats, humans and neural networks discriminate amplitude-modulated FM echoes [abstract]

L. A. Miller and D. A. Helweg (2002). Bats, humans and neural networks discriminate amplitude-modulated FM echoes [abstract]. Bioacoustics, Volume 13 (1): 86

 

Abstract: 

The echoes received by bats can be amplitude modulated by movements and by the structural properties of the target. Amplitude modulated echoes will have a different envelope relative to the emitted sonar pulse. The envelope contains considerable information about the target if frequency modulated (FM) signals are used. In theory, an FM bat could identify a target using information in the envelope of the echo. Behavioural thresholds for detection of amplitude modulation were obtained in the Big Brown Bat Eptesicus fuscus. Bats learned to discriminate amplitude modulated echoes from non- modulated echoes with some difficulty. Performance was best at AM rates of 3 and 6 kHz, and fell off at 1 and 12 kHz. It was unclear what cues the bat used to detect AM. Two models were explored. One, human listeners were asked to perform the same discrimination using time-stretched digitised echoes. Two, backpropagation networks were trained to classify digitised echo spectra as either modulated or non-modulated, while systematically varying spectral content. The performance of the bats diverged from that of the human listeners and the backpropagation networks. Results for the humans and networks demonstrated an upward shift in the best range to 6 and 9 kHz. The acoustic basis of the bat's performance remains highly intriguing and deserving of further attention. lmplications will be discussed in light of echolocation behaviour and ecology of Eptesicus [Supported by grants from the Danish Research Council to L. Miller and University of Auckland Research Council to D Helweg].