Detection of targets colocalized in clutter by big brown bats (Eptesicus fuscus)

Abstract

Echolocating big brown bats (Eptesicus fuscus) frequently catch insects during aerial pursuits in open spaces, but they also capture prey swarming on vegetation, and from substrates. To evaluate perception of targets on cluttered surfaces, big brown bats were trained in a two-alternative forced-choice task to locate a target, varying in height, that was embedded partway in holes (clutter) cut in a foam surface. The holes were colocalized with the possible positions of the target at distances ranging from 25 to 35 cm. For successful perception of the target, the bat had to detect the echoes contributed by the target in the same time window that contained echoes from the clutter. Performance was assessed in terms of target reflective strength relative to clutter strength in the same time window. The bats detected the target whenever the target strength was greater than 1-2 dB above the clutter.

Figure 1

(A) Diagram of the two-alternative forced-choice detection task. The bat was placed on the Y-shaped platform and trained to detect a target consisting of two small cylinders located on the right or left side. The cylinders were 5 cm apart and were embedded in holes cut into the surface of the 2.5 cm thick foam disks 90 cm in diameter. They varied in height by 12, 8, 4, or 2 mm above the surface of the foam (see inset in B). To receive food reward, the bat walked down (arrow) the side of the platform that corresponded to the target. The target’s cylinders occupied two holes on the left (as indicated in B) or the right (alternated pseudorandomly), and they reflected echoes whose strength depended on the cylinders’ protruding height. The other holes were empty from one trial to the next, but they generated their own reflections (inset in C) to compete with echoes from the cylinders and serve as clutter. For successful detection, the bat had to discriminate the reflections of the cylinders from these added clutter reflections. The experiment measured the bat’s detection performance while the strength of the target’s reflections was decreased in several steps by reducing the protruding height of the cylinders until target strength was similar to the clutter.

Figure 2

Echoes from the clutter (holes) and target (dipole cylinders) were measured by projecting a 1 ms long FM signal (sweeping from 110 to 15 kHz), similar to the bat’s echolocation emissions, from a 2 cm electrostatic loudspeaker and recording the echoes with two condenser microphones. To make the dipole echo measurements, the cylinders were placed in the foam holes with a specific protrusion height (12, 8, 4, or 2 mm) and the entire scene was ensonified to generate reflections.

Figure 3

(A) Spectrograms of reflections from the target at an orientation of 50° [see Fig. 1B] for different amounts of protrusion of the cylinders above the surface of the foam disk. The double sweeps in these spectrograms trace the separate reflections from the cylinders, which are oriented so their echoes arrive about 180 μs apart (see C for reference). (B) Output of matched filter (cross correlation receiver with full-wave-rectified plots) for the same series of reflections as in A. Red and blue curves trace outputs from left and right microphones (see Fig. 2). The twin peaks in the cross covariance curves register the separate reflections from the cylinders, while their heights have the same scale to indicate relative target strength. (C) Spectrogram for the specular reflection from a flat target oriented perpendicular to the sound from the loudspeaker (Fig. 2). The sweep pattern in C is for reference to the spectrograms in A, which show two closely spaced, overlapping sweeps from the two cylinders. The plots are labeled with the target’s protrusion height (12, 8, 4, or 2 mm and empty holes). The empty holes and the blank foam surface produce acoustic scattering (clutter) across the same time window as the echoes from the target. (D) Performance (percent correct responses) of four big brown bats in detection task with different amounts of cylinder protrusion. Performance closely mirrors visibility of the FM sweeps in the spectrograms (A) or the prominence of the peaks in the cross covariance curves (B).

Figure 4

Graph showing the mean performance of all four bats for different increments in target strength expressed in decibels relative to the target strength of the holes alone. Target strength is the energy reflected in a 250 μs integration-time window around each cylinder’s reflection delay [see Fig. 3B], expressed in decibels relative to the energy reflected by the holes in the same time window. The bat’s performance was significantly greater than chance (p<0.05) for echo increments greater than 1–2 dB.