mechanism, training and limits of human echolocation
TRANSCRIPT
Mechanism, training and limits of human echolocationLéopold Kritly1,3,David Pelegrin Garcia 2,Antoine Gademer 3, Christ Glorieux 4, Monika Rychtarikova1,5
1 Faculty of Architecture - KU Leuven 2 Widex A/S3 Institute of Systems and Electronics - EPF Montpellier 4 Laboratory of Acoustics - KU Leuven 5 Faculty of Civil Engineering - STU Bratislava
Presentation outlines
Introduction
Mechanism
Training
Limits
Perspectives
Basis
Usage
Operating cues
Introduction
Stick
Environmental noise
External help
Echolocation
20 Minutes, Phlessons,
Cité des Sciences et de l’Industrie
Introduction
V ≈ 299.8 km/s V ≈ 342 m/sX 1000
Processing of environment perception
Immediate Sound reflections arrive not at the same time
Mechanism
Ability to use sounds to locate and identify the objects aroundWhat is echolocation ?
Active Passive
Human head
Self sound emission
Direct Sound
Obstacle reflection
Propagation through environment
Echoes
Environment noise
Human head
Obstacle reflectionDirect Sound
Echoes
Propagation through environment
Pelegrin-Garcia D., Rychtarikova M., Zelem L., Chmelik V., Kritly L., Glorieux C. (2017). Can humans use echolocation to hear the difference between different kinds of walls ? ASA, 2669-2.
MechanismActive echolocation: Self sound emissions
Emission Tongue click Whistles Hisses Speech
Spectrum content Broad band Tonal Broad band Speech range
[200 : 4000] Hz
Intensity High Small
Duration Impulsive Long
Usage Polyvalent Short range & sound coloration detectionsRequired silent environment
Most used
MechanismWhat could we detect through echolocation ?
• Location
• Size
• Intrinsic features
• Shape
• Texture
Information about obstacle:
MechanismObstacle location - Distance
Time of propagation
RDLD = Lrefl – Ldir
t
t
Left ear
Right ear
p
Reflected to Direct Level Difference (RDLD)
MechanismObstacle location - Orientation / Localization
Oral-Binaural Room Impulse Response (OBRIR)
t
t
Left ear
Right ear
p Interaural delay
GitHub – CSSonar [Online] – Josh Lind – 2015
MechanismFeatures from objects recognition
Changes in spectrum (emission / emission + echoes )
Kolarik, A. J., Cirstea, S., Pardhan, S., & Moore, B. C. J. (2014). A summary of research investivating echolocation abilities of blind and sighted humans. Hearing Research, 310, 60-68. Compact Dynamics – Acoustical Material and Acoustics [Online] - http://www.12345w.xyz/compactdynamics-com/encyc-acoustics-0.html
Carpet
Fabrics
Wood
Plexiglas
C.E. Rice, S.H. Feinstein & R.J. Schusterman (1965). J.Exp.Psychol. 70, 246
MechanismMinimum detectable target size
xxxxxxxx
MechanismMinimum detectable target size
Audibility Thresholds of a Sound Reflection in a Classical Human Echolocation ExperimentPelegrin-Garcia, Rychtarikova, Glorieux - Acta Acustica United With Acustica - Vol 102 (2016) 530-539
Increase of diameter sizeo reflected level decreaseso low frequencies are
reflected more efficiently
• OBRIR meas. on disks of 4 different sizeo 5.6 cm, 11.6 cm, 23.8 cm, 38.6 cm,
• 5 different distances(case of Rice’s study)
o 0.5m, 1m, 1.5m, 2m, 2.5 m
MechanismConstitution shape recognition
Mapping of echoic informationDifficulty to recognize Shape
Easy
Medium
Hard
Kolarik, A. J., Cirstea, S., Pardhan, S., & Moore, B. C. J. (2014). A summary of research investivatingecholocation abilities of blind and sighted humans. Hearing Research, 310, 60-68.
Mesh space
• interval that contains between 0.5 and 99.5% of the energy
• thick bars contains the central 75% of the energy
• dots show the central time (same energy before and after)
MechanismTexture recognition
Reflected to Direct Level Difference (RDLD)
Temporal spread of reflections
Pelegrin-Garcia D., Rychtarikova M., Zelem L., Chmelik V., Kritly L., Glorieux C. (2017). Can humans use echolocation to hear the difference between different kinds of walls ? ASA, 2669-2.
MechanismA powerful mechanism
Cycle
Training – KU Leuven facility
Hike in nature
Ski
Seth Freeman – 2011,Wowscience – Brighter Life Experiences for Blind People with the use of echolocation [Online] - 2017,
Youtube / Poptech – Daniel Kish’s echolocation in action [Online] – 2011
• Goals of using a VR systemo Now: Better understand the underlying perceptual mechanisms o Future: Help in echolocation training and familiarization with
unknown environments
Training
VR Audioprocessor
t
t
Left ear
Right ear
p
Oral-Binaural Room Impulse Response (OBRIR)Emulation with acoustic Virtual Reality (VR)
Training
1. Audibility of sound reflections
2. Discrimination - localization
3. Feature recognition (size, shape, texture…)
4. Complex tasks – environment identification, obstacle avoidance
5. Use in daily life – integration with other mobility techniques
Sens
atio
n
Perc
eptio
n
Cog
nitio
n
The 5 layers of human echolocation
Fundamental aspects
ResearchInterest
LimitsPerformance in complex environments
o Reverberant
o Noisy
Improvable buildings design for echolocators
Wikipedia – St Andrew Church, PxHere,
Deviantart – Isstatu1
PerspectivesUnderstanding echolocation
in architectural context
Design of acoustic markers Discrimination of features
necessary for echolocation
Training of people who became
blind during their lifetime
Thank you for your attention
• Results averaged for the two ears:
Comparing real vs virtual experimentsReflected-to-direct level difference (RDLD)
• Increase of diameter sizeo reflected level decreaseso low frequencies are
reflected more efficiently
• the further one go the lower the the level of reflection is
Comparing real vs virtual experimentsReflected-to-direct level difference (RDLD)
• Single number RDLDo Frequency weighting: “typical oral click” + inverted loudness level curve
from ISO-226:2003o Uncertainty of ~3 dB due to reference click choice and anatomical
differences
Comparing real vs virtual experimentsReflected-to-direct level difference (RDLD)
• Rice et al.’s (1965) thresholds revised
• RDLD Threshold: -20 dB independent of distance
• +/- 3 dB spread across participants
• Use as reference performance
That’s how we can translate thresholds vs solid angle.
Comparing real vs virtual experimentsReflected-to-direct level difference (RDLD)
That’s how we can translate thresholds vs solid angle.
RDLD, as a function of the distance to the obstacle =the equivalent time delay of the reflection
black = average
gray symbols = individual results
error bars ± 1 standard deviation
horizontal gray bold lineaverage RDLD correspondingto a constant solid anglecovered by the target
Sensitivity to a single reflectionResults
• Click durationo Exp. I: Pre-recorded sounds 3.1 mso Exp. II: Self-generated sounds 12 ms
RDLD thresholds lower with distance (but also do reflection levels!)
Longer clicks worse thresholdsVariability in “click” production more spread in thresholds
Masking is non-linear louder clicks result in better thresholds
/s/-like sounds only effective at short distances
Thresholds for blind echolocators ~ 10 dB better at short distances(compare only black filled symbols)• Better signals?• More trained to detect coloration?• Inhibition from direct sound?
RDLD of stimuli• conditions at 1.5m are above thresholds,
except for the wall 7 and 12• conditions at 10 m are much further from threshold
than at 1.5 m.