25.05.2014 Views

Bat Echolocation Researc h - Bat Conservation International

Bat Echolocation Researc h - Bat Conservation International

Bat Echolocation Researc h - Bat Conservation International

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ZERO-CROSSINGS ANALYSIS FOR BAT IDENTIFICATION:<br />

AN OVERVIEW<br />

CHRIS CORBEN<br />

In this paper, I present an overview of the effectiveness of Zero-crossings Analysis (ZCA) for the identification of<br />

free-flying bats, based on 17 years of experience using the Anabat system. My opinions are based on intimate<br />

involvement with ZCA, as the original designer of the hardware and software used by Anabat, and also as one<br />

who routinely uses Anabat as a major component of field surveys for bats. I explain how the system works and<br />

how it compares with other approaches to the general problem of making echolocation calls available to humans<br />

for bat identification. I argue that ZCA has numerous advantages over other approaches, especially for passive<br />

monitoring. I also offer my views on a number of general issues pertaining to the use of acoustic monitoring for<br />

bats.<br />

Key words: Anabat, bats, echolocation, Fast Fourier Transform, frequency division, heterodyne, spectral analysis, zero-crossings<br />

Correspondent: corben@hoarybat.com<br />

INTRODUCTION<br />

I became involved with bat acoustics in the mid-1980s<br />

when I saw the chance to combine my passions for electronics<br />

and wildlife in the design of a practical bat-detection<br />

system. I was intrigued by papers about the use of<br />

bat echolocation calls for species identification, but I<br />

found that equipment then available for making use of<br />

ultrasonic signals was too expensive for my budget. I realized<br />

that recent advances in computer hardware would<br />

make it possible to find other solutions. Consequently, I<br />

resolved to develop a less-expensive system, better tailored<br />

to the specific needs of bat acoustic identification.<br />

I designed both hardware and software for a system based<br />

on Zero-Crossings Analysis (ZCA), subsequently developed<br />

as a commercial product called Anabat (Titley Electronics,<br />

Australia, www.titley.com.au). In this paper, I use<br />

the terms Anabat and ZCA interchangeably because<br />

Anabat is the implementation of ZCA with which I am<br />

most familiar, and which is most widely used for monitoring<br />

bats, at least in Australia and North America.<br />

In recent years, my main employment has been to survey<br />

bats for various purposes, and to this end I have made<br />

extensive use of ZCA, as implemented by Anabat. I find<br />

this system to be effective for both field identification of<br />

bats and passive monitoring of bat activity. Using Anabat<br />

has allowed me to treat bats as field-observable animals,<br />

in much the same way as birds or frogs.<br />

For the symposium, I was originally asked to write on<br />

the subject of frequency division, but this is a relatively<br />

minor part of the whole system, so I mostly address ZCA<br />

and its strengths. To do so requires comparing ZCA with<br />

spectral analysis, which is the principal alternative<br />

means of visually displaying bat calls. I argue that ZCA<br />

is a more practical and effective approach to echolocation-call<br />

analysis for species-identification purposes. I<br />

have extensive experience with spectral analysis, mainly<br />

for analyzing frog calls, though I have also used it to<br />

Section 3: Ultrasound Species Identification<br />

analyze vocalizations of birds and bats, and other<br />

sounds. I greatly admire the concept of spectral analysis,<br />

and I think of the Fast Fourier Transform (FFT) as one of<br />

the most elegant mathematical inventions. The sounds<br />

of most animals are best analyzed using spectral analysis,<br />

but for identifying free-flying bats, this approach is technological<br />

overkill because bat echolocation calls are simple<br />

signals well suited to analysis by ZCA. It is therefore<br />

appropriate to make use of the many advantages ZCA<br />

offers in cost and efficiency. In my view, spectral analysis<br />

does not provide clear advantages that offset its practical<br />

liabilities for this purpose.<br />

HOW DO FREQUENCY DIVISION AND ZCA WORK?<br />

One way to determine the frequency of a signal is to<br />

count how many cycles, or vibrations, occur in a given<br />

time interval. This works well when dealing with continuous<br />

signals of constant frequency. However, bat<br />

echolocation call frequencies typically change rapidly,<br />

so another approach is required. Instead of counting<br />

cycles, the duration of each individual cycle can be measured<br />

and the frequency computed from its reciprocal to<br />

produce a graph of frequency against time. This is the<br />

basis of ZCA.<br />

In practice, it is often advantageous to first derive a<br />

second signal, which produces one cycle for each group<br />

of several cycles from the original signal. This is the process<br />

of frequency division. Using it has the advantage<br />

that the frequency-divided signal is downshifted in frequency<br />

by a constant factor, the division ratio, making it<br />

audible to humans. Another advantage is that a frequency-divided<br />

signal averages out small fluctuations, giving<br />

a smoother representation, better suited to the level of<br />

detail needed to see the underlying structure of a bat<br />

call. A lower division ratio displays more points in the<br />

frequency display but at the cost of increased storage<br />

requirements and processing power, and if the extra<br />

95

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!