Active echolocation in bats and humans

1

• Research questions
• Can blindfolded humans can use echoes to
  determine the left-right position of a reflective
  board?
• Based on acoustic analysis, what cues might be
  available to do this and at a range of distances
  and board orientations?
• Based on listening tests, what cues are used
  and how important is distance and board
  orientation?

Active echolocation in bats and humans This
modality is used by animals such as bats to
navigate and find food. There has been little
study of human echolocation despite some
charities training blind subjects to use it. Our
research is based on a paradigm described by
Despres et al (2005) in which blindfolded
subjects could identify whether a board was to
their left or right, using acoustic cues from a
speaker below the head.
Free field board localisation The Depres et al
was extended to include (1) further steps to
ensure judgement was not due to auditory cues
unrelated to board reflections (2) no-sound
trials (3) clicks. Some subjects had
above-chance performance but there was much
inter-subject variation. No significant
difference between click and white noise stimuli.
See Figure 1.
Listening tests Objective results were
corroborated by listening tests in which subjects
were presented with broadband stimuli convolved
with the measured binaural impulse responses.
Some results are shown in figure 4.
Board right or left?
Fig. 1. Left panel illustration of human
experiment conducted in an anechoic chamber.
Stimulus (60 dB SPL) is produced from the speaker
(dark) directly below subjects chin. Subjects
report whether board is to the left or right.
Side speakers masked sounds made while moving
the board. Right panel results of an experiment
using 12 inexperienced normal-hearing blindfolded
subjects. A common psychoacoustic measure,
d-prime (related to correct zero is chance) is
used to measure detectability in three stimulus
conditions 10-s-long series of bursts of
broadband noise and clicks and no sound.
Fig 4. Results of headphone experiment involving
13 naïve normal-hearing listeners. Noise
signals (400-ms-long) of different bandwidths
were convolved with impulse responses from the
board (measured at KEMARs ears) in various
stimulus and board conditions. The final stimulus
with broadband noise and the board flat, central
and at0.9 m was played at 65 dB SPL to both
ears. Single interval left vs. right task,
with trial-by-trial feedback.Noise was low- or
high-passed filtered at 2 kHz prior to
convolution. Mean sensitivity across listeners is
shown with error bars illustrating 95 confidence
intervals.
Measurement of potential cues Impulse responses
taken in an anechoic chamber at the ears of
KEMAR were obtained as shown in figure 2.
Speaker localisation rig A rig designed to mimic
the signal and emission that would reach the
subject was used to investigate the effects of
duration and using a click train rather than a
single emission for signals of various durations.
See figure 5.
Fig. 2. Left panel method for measuring impulse
recordings when a 0.55 cm2 reflective board was
placed between 40 cm and 400 cm from a KEMAR
head. The centre of the loudspeakers driver was
0.25m below KEMARs ears. Right panel Impulse
responses were measured with the centre of the
board at 17 from the centre of KEMARs head,
either when flat or at an angle, using the MLS
technique at 88.89kHz sampling rate with the
cut-off of the anti-aliasing filter set to 20kHz.
Recordings obtained Analysis indicates the
existence of potentially useful ILD cues. ITDs
are not apparent with the board flat but are with
the board at an angle. See figure 3. , Monaural
cues are unlikely to be of use to untrained
listeners
Fig. 5. Left panel illustration of rig. 11
inexperienced subjects placed their heads in the
centre of a speaker array. A central speaker
simulated a 54 dB SPL emission and side speakers
simulated a reflection at 90 cm. Subjects stated
whether the emission was from the left or right.
Feedback was given. Right panel mean sensitivity
at 17. Interaction between duration (ms) and
stimulus type (1 vs. 10 clicks) was significant.
Duration had no significant effect on d. Post
hoc t tests showed that at each duration,
performance of a click train was significantly
greater than for a single click.

• Conclusions and further work
• The ability to echolocate depends greatly on
  task and signal design.
• In certain conditions, humans can use
  echolocation cues to determine object position
  though the cues used may depend on object angle.
• Click trains appear to give more information
  than a single emission though the effects of
  other temporal features of the signal are yet to
  be investigated.
• Comparing situations with and without emission
  and signal overlap in time may give some insight
  into the workings of the precedence effect.
• The board paradigm allows formulation of an
  experimental protocol for conducting more
  informative experiments in human echolocation.

Fig 3. impulse recordings for an angled board
positioned 17 to the left at 90 cm showing
emission (left) and reflection (right). Upper
panel left ear response. Lower pane right ear
response. Time and level differences between the
recordings are apparent.
Acknowledgements The authors acknowledge the
support of RCUK through its Basic Technology
Programme. David Edwards was supported by a EPSRC
studentship grant and Hannah Holmes by an EPSRC
Vacation Bursary.
Reference Despres, O, Candas, V and Dufour, A.
(2005) Enhanced sensitivity to echo cues in blind
subjects. Experimental Brain Research, 165,
515-519.