Determining the Binaural Signals in Bat Echolocation Timos
Total Page:16
File Type:pdf, Size:1020Kb
Advances in Science and Technology Vol. 58 (2008) pp 97-102 online at http://www.scientific.net © (2008) Trans Tech Publications, Switzerland Online available since 2008/Sep/02 Determining the binaural signals in bat echolocation Timos Papadopoulos 1,a , Robert Allen 1,b and Stephen Haynes 1,c 1ISVR, University of Southampton, SO17 1BJ, UK [email protected] [email protected] c [email protected] Keywords: Bat echolocation, bioacoustics modelling, bio-inspiration. Abstract. Echolocating bats are known to outperform manmade systems in the tasks of autonomous navigation and object detection and classification, especially when size, power and computational complexity requirements are considered. As a result, the individual physical mechanisms and processes involved in echolocation (types of signals used, properties of the emission mechanism, echoes created in different echolocation tasks, receptor characteristics as well as the bat’s auditory system) have received significant attention as a possible source of bio-inspiration. However, not much attention has been drawn to optimisations that may arise as a combined effect of the above mechanisms. Of key importance in such an investigation would be the knowledge of the binaural signals generated in real echolocation tasks as those are the actual input signals utilised by the bat’s auditory system. The direct measurement of these signals is severely restricted by the very small size of most bat species. We describe the development of an experimental facility that combines the measurement and modelling of the aforementioned subsystems for the determination of the binaural signals associated with echolocation. We present initial measurement results and compare them with analytical modelling predictions. Introduction The drive to improve acoustic measurement and imaging systems leads to ever increasing frequencies and energy yet the resolution and characterisation capabilities lag way behind those of animals. Echolocating bats and dolphins, for example, exhibit impressive acoustic performance. Some bats, for example, are capable of discriminating jittered time pulses down to 10ns at frequencies around 80kHz with a 40kHz bandwidth [1], a resolution several orders of magnitude beyond current technology. In addition, bats have been filmed during flight avoiding closely spaced wires of less than 1mm in diameter [2] which demonstrates an impressive capability in detecting and localising sub-wavelength targets. Material discrimination through acoustic signalling by dolphins again significantly outperforms current detection systems and yet this ability would provide a step change in acoustic technology if the mechanisms and signalling strategies could be understood. Animals, however, have evolved over many years and their acoustic systems have been developed to achieve specific tasks, environments and prey. Engineering systems, on the other hand, usually have to operate within a much wider envelope but clearly have much to learn from biological systems. Current understanding of the key acoustic signalling strategies and cues used by echolocating animals is, however, still largely incomplete and unravelling the mysteries requires a multidisciplinary team approach. This is the view taken by the Biologically Inspired Acoustic Systems (or BIAS) Consortium which brings together expertise in animal acoustics, signal processing theory, transducer design and experimental testing, together with specialists in key application areas. A key word in the consortium name is inspired . The intention is not to attempt to mimic the natural systems but to be inspired by their capabilities and to identify how they function in order to generate new acoustic technologies for the future. The work presented here focuses on the study of bat echolocation and the relevant physical acoustics modelling in air. Our aim is to investigate the possibilities and limitations involved in the theoretical design and experimental implementation of an acoustics model that can predict the All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 152.78.202.118-14/10/08,12:02:44) 98 Mining Smartness from Nature acoustic pressure signals generated at the ears of the bat during echolocation. The motivation behind such a goal has been discussed in previous studies [3,4] but the first real efforts to experimentally realise it started appearing only very recently [5,6,7]. Furthermore, even in those studies, the recorded acoustic pressure signals were obtained by a microphone mounted on the body of the bat rather than inside its ear. This practically means that such measurements do not include the contribution of the bat’s head and pinnae in the reception mechanism. Obviously, any effort to directly measure the signals reaching the bat’s ears during echolocation would be severely hindered by the very small dimensions of the bat’s ear canal which compared to the finite size of the microphone and transmission apparatus makes such a goal practically unattainable. Furthermore, such direct measurement methods require a separate individual measurement for each combination of different echolocation signal, target type and geometry as well as the varying shapes and sizes of head and pinnae in different bat species. Aiming to overcome that, in a previous paper [8], we described the theoretical basis of a method that can be used for the determination of the bat binaural signals during echolocation. By virtue of modelling the echolocation process as three linear and time-invariant subsystems corresponding to the emission, echo formation and reception mechanisms, the method can in principle combine separately measured or otherwise acquired models of each subsystem to predict the binaural signals corresponding to a variety of possible combinations. For the results presented in this paper we devise a simplified artificial echolocation- type experiment that allows us to directly measure the impulse response between the input to a source and the signal captured by a microphone mimicking the mouth and ear on a bat’s head. We then compare this to the impulse response predicted by the combination of the measured the models corresponding to the three subsystems described above in order to obtain a first estimate of the limits of applicability of the method. Experimental method Our experimental setup is described in the following figures. As is shown in part (a) of Fig. 1, a small loudspeaker cabinet of approximately (20cm) ×(20cm) ×(20cm) dimensions is fitted with a piezoelectric tweeter source in its front face and an audio microphone at its side face thus imitating a (much larger than normal) “Bat head”. A target is mounted at a distance L in front of the “head” and the impulse response, denoted here with A(n), is measured between the electrical input to the source and the microphone output pear (n) using a maximum-length sequence excitation stimulus. Knowledge of this response for a given echo generating object and a given geometry allows the determination of the signal pear (n) corresponding to the arbitrary input signal x(n) by means of convolution: pear (n)= A(n)* x(n). However, as was described above, the direct measurement of A(n) in real echolocation situations is problematic. Our objective is then to obtain a model of A(n) using the measurable quantities described in the following figures. (a) (b) target “Bat head” “Bat head” Measurement microphone pout , pecho L1 L2 L1 δ(n) pear (n) L δ(n) pout (n) Fig. 1 Part (a): Schematic of the acoustics model. A delta pulse input to the source results in the outgoing sound field and the pressure signal pout at a distance L1. The sound field is backscattered by the target and the signal pecho is created back at the same point. The backscattered sound field is diffracted by the bat’s head and pinnae resulting in the signals pear at the bat’s ears. Part (b): Direct measurement of the signal pout . Advances in Science and Technology Vol. 58 99 The first of these quantities is the acoustic pressure signal generated at a given distance along the line between the bat and the echo generating object, due to the echolocation emission. Databases of such signals can be obtained either by field measurement [9], or by the elicitation of echolocation calls in the lab [10] or by analytical models of the directionality of the emission mechanism [11]. For the results presented here, we measured the impulse response B(n) between the input to the source mimicking the bat’s mouth and an audio microphone at a distance L1 directly in front as is shown in part (b) of Fig. 1. Note that the sensitivity of the microphone will appear as a free scaling parameter in the measured impulse response, the significance of which becomes apparent in the results presented in the next section. (a) (b) Measurement Measurement target “Bat head” Measurement source microphone source L2 δ(n) p p p (n) δ(n) out , echo ear Measurement Measurement Measurement Measurement microphone source source microphone δ(n) L1 p (n) p (n) δ(n) ff ff Fig. 2 Part (a): Schematic of the backscattering model measurement. Part (b): Schematic of the HRTF model measurement. The second subsystem we are seeking to model is the acoustic backscattering that creates the echo signal that travels back to the bat. A detailed description of our proposed modelling for this subsystem can be found in [12] and a schematic description is given in part (a) of Fig. 2. Very briefly, the modelling consists of measuring the impulse response C1(n) between the input to a measurement source and the output at a microphone positioned at a distance L2 from the target such that the sum of L1 (see part (b) of Fig. 1) and is equal to the distance L between the bat and the target (see part (a) of Fig.