Bat Pulses at High Resolution

Bat Pulses at High Resolution
Seeing bat pulses in fine detail with the LPC Spectrogram
Neil J Boucher , Michihiro Jinnai
1
2
SoundID, Maleny, Australia, 2Nagoya Women's University, Japan
1
Introduction
Until recently the FFT has been the main tool that researchers have used
to examine the structure of bat pulses. Unfortunately the FFT is a very
blunt tool indeed and when the same pulses are examined under a more
powerful spectrum analyser a whole new world of detail opens up. Here
we look at some examples for bats.
Objectives
We have been using a high resolution spectrum analyser (the LPC) for
sound recogniion. We applied it to some bioacoustic sounds to see how
much they differed from the much more basic FFT. We found a lot of differences and we would like to share some of what we found.
The LPC image of the Mydau pulse.
Methods
We used the SoundID Cluster Analysis module to get the LPC graphics
and Cool Edit for the FFT graphics.
Results
Here you will see the FFT of the pulse, which suggests a wide-band descending pulse. The LPC however shows a narrow band pulse wobbling
as it descends ( technically, wobulating).
The FFT image of the Mydau = Myotis daubentoni
Zooming into approximately 20% of the pulse (0.03 seconds which is typical of the rest of the pulse), we notice in the WAV view (lower part of the
screen) there is a modulation envelope for the pulse. Additionally the average frequency is slowly decreasing. This is an example of a wave packet,
which by definition has an indeterminate frequency (we call this wobulation). In the spectrum part of this image you can see the pseudo-random
wobulation.
Now we look at another bat pulse that was described as being either a
Mydau or a pipi-nyno (they are mixed in the samples, Pipi = Pipistrellus
pipistrellus, Nyno = Nyctalus noctula ). There is nothing much special
about it seen in the FFT. But look at the LPC below.
Here in the LPC view we see that it wobulates as it falls in frequency but at
the lower end of the frequency (and the last envelope) it wobulates more
widely. It almost appears to have two frequencies but if you look carefully it is merely fluctuating. Dramatic changes in the structure of signals
separated by different envelopes has been seen by us in birds also. Additionally if we look at a small section of the WAV file below from the
same call we see rather severe distortion in the individual pulses. This
may be something that the bat is doing (some birds do this at times to
produce harmonics) or it could be something as simple as distortion in
the recorder. But there is no doubt that the distortion, however caused,
will lead to some indeterminancy in the frequency.
Conclusions
This study has been done to show what is possible with high resolution
tools and the serious shortcomings of the FFT. There is a lot more to be
learnt (and almost as much to be unlearnt!).