"From
ITU-T Workshop on
Speech to Audio: bandwidth extension,
binaural perception"
Lannion, France, 10-12 September 2008
Acoustic Impedance Characteristics of
Artificial Ears For Telephonometric Use
Gaëtan Lorho,
Nokia Corporation
Lannion, France, 10-12 September 2008
International
Telecommunication
Union
Overview
Artificial ears for telephonometry
Measurement campaign
Acoustic impedance testing
Impedance measurement results:
Human ears
Artificial ears
Human vs. artificial ears
Conclusions
Paper available on the ITU-T workshop page
Lannion, France, 10-12 September 2008
(www.itu.int/ITU-T/worksem/speechaudio)
International
Telecommunication
Union
2
Artificial Ears for Telephonometry
Analog of human
ear for objective
acoustic
measurement
Efficient and
repeatable way of
predicting acoustic
performance of a
handset in final
usage case
Lannion, France, 10-12 September 2008
Standardization
within industry:
ITU-T Rec. P.57:
Artificial Ears
Ears should
provide“[…] an overall
acoustic impedance
similar to that of the
average human ear
over a given frequency
band”
International
Telecommunication
Union
3
Artificial Ears for Telephonometry
ITU-T Rec. P.57
Type 3.3 and 3.4
become industry
de facto
Primarily for use
with a HATS
ITU-T Rec. P.58
“Head and torso
simulator for
telephonometry”
Lannion, France, 10-12 September 2008
Type 3 - IEC 60711
occluded ear
simulator with canal
extension
terminated in a:
Type 3.3 pinna
simulator
(anatomically shaped)
Type 3.4 pinna
simulator (simplified)
International
Telecommunication
Union
4
Measurement Campaign
ITU-T SG12
responsible for
development of
Rec. P.57
Recent effort to
benchmark Type 3
ears against the
“average” acoustic
impedance of
human ears
Lannion, France, 10-12 September 2008
Specific focus on:
Mobile phone in
‘hand-portable’ mode
Measurement for a
large number and
range of human users
Wide-band
frequency range
Round-robin
campaign (multiple
industry participants)
International
Telecommunication
Union
5
Acoustic Impedance Testing
Mobile phone-like
impedance probe
provided by Brüel
& Kjær
Lannion, France, 10-12 September 2008
Frequencydependent
impedance measure
close to the ECRP
International
Telecommunication
Union
6
Acoustic Impedance Testing
Human measures:
5 contributing (independent) laboratories
Subject gender and age demographic
considered
106 subjects total measured
2 separate handset application force measures
made per subject
‘Normal’ application force (inferred from
placement in silent condition)
‘Firm’ application force (inferred from
placement in lab. simulated noise field)
Impedance measure at each ISO R40 1/12th
oct. between 0.2-8kHz for each test case
Lannion, France, 10-12 September 2008
International
Telecommunication
Union
7
Acoustic Impedance Testing
Artificial ear measures:
Commercially available artificial ears tested
with same probe by each manufacturer
Brüel & Kjær Type 3.3 ear
HEAD acoustics Type 3.4 ear
Measurement on HATS at ‘standard’ position
Application forces between 2 and 18N
increasing by 2N steps
Impedance measure at each ISO R40 1/12th
oct. between 0.2-8kHz for each test case
Lannion, France, 10-12 September 2008
International
Telecommunication
Union
8
Testing Results: Human Ears
Statistical analysis for each individual
1/12th octave band
240
All human measurements - "Firm" application force
Human meas. - Firm ap. force - conf int. & stand dev.
240
220
Za*w (Magnitude)
Za*w (Magnitude)
220
200
180
180
160 2
10
200
10
3
Frequency (Hz)
Lannion, France, 10-12 September 2008
10
4
160 2
10
3
4
10
10
Frequency (Hz)
International
Telecommunication
Union
9
Testing Results: Human Ears
Illustration of the two most important
factors: Application force and Gender
Average of human meas. per gender
Average of human meas. - Normal versus Firm ap. force
220
Za*w (Magnitude)
Za*w (Magnitude)
220
200
200
180
Blue: Normal application force
Red: Firm application force
10
2
3
10
Frequency (Hz)
Lannion, France, 10-12 September 2008
Black: Male
Blue: Female
180
10
4
10
2
10
3
4
10
Frequency (Hz)
International
Telecommunication
10
Union
Testing Results: Human Ears
Detection of impedance response extrema
“Structural modeling” of measures
Impedance extrema of human meas. - Normal appl. force
Structural description of human meas. - Normal appl. force
240
Three first maxima
of human responses
Mean of extrema and
ellipse covering 95% of
the individual extrema
Structural mean
Za*w (Magnitude)
Za*w (Magnitude)
220
200
220
200
180
3 individual
human responses
160 2
10
180
Two first minima
of human responses
3
10
Frequency (Hz)
Lannion, France, 10-12 September 2008
4
10
2
10
3
10
4
10
Frequency (Hz)
International
Telecommunication
11
Union
Testing Results: Human Ears
Comparative analysis of structural means
Structural model of human meas. - Normal appl. force
240
240
Structural mean
220
Za*w (Magnitude)
Za*w (Magnitude)
95% confidence ellipse
of the extremum mean
200
Arithmetic mean
and 95% conf. interv.
Normal appl. force
180
10
Structural model of hum. meas. - Norm. vs. firm appl. force
2
10
3
Frequency (Hz)
Lannion, France, 10-12 September 2008
220
200
180
10
4
10
Blue: Normal application force
Red: Firm application force
2
3
10
10
4
Frequency (Hz)
International
Telecommunication
12
Union
Testing Results: Artificial Ears
Individual 1/12th oct. response for each
artificial ear type per application force
240
B&K - HATS meas. - type 3.3 - 9 pressure levels
240
220
Za*w (Magnitude)
Za*w (Magnitude)
220
200
180
From low (2N) to
high pressure (18N)
160
140 2
10
HA - HATS meas. - type 3.4 - 9 pressure levels
200
From low (2N) to
high pressure (18N)
180
160
3
10
Frequency (Hz)
Lannion, France, 10-12 September 2008
4
10
140 2
10
3
10
4
10
Frequency (Hz)
International
Telecommunication
13
Union
Human vs. artificial ears
Human individual 1/12th octave band
CI95% vs. artificial ears types
Average of hum. meas. at Norm. and Firm force versus 3.3
Average of hum. meas. at Norm. and Firm force versus 3.4
220
Za*w (Magnitude)
Za*w (Magnitude)
220
200
From low (2N) to
high pressure (18N)
180
160 2
10
Blue: Normal application force
Red: Firm application force
3
10
Frequency (Hz)
Lannion, France, 10-12 September 2008
10
4
200
From low (2N) to
high pressure (18N)
180
160 2
10
Blue: Normal application force
Red: Firm application force
3
10
10
4
Frequency (Hz)
International
Telecommunication
14
Union
Human vs. artificial ears
Human structural mean vs. artificial ears
types
Structural model of hum. meas. at Firm force versus 3.3
Mean of extrema and
ellipse covering 95% of
the individual extrema
Structural model of hum. meas. at Firm force versus 3.4
240
Structural mean
Firm appl. force
220
Za*w (Magnitude)
Za*w (Magnitude)
240
200
Mean of extrema and
ellipse covering 95% of
the individual extrema
220
200
From low (2N) to
high pressure (18N)
From low (2N) to
high pressure (18N)
180
10
Structural mean
Firm appl. force
180
2
3
10
Frequency (Hz)
Lannion, France, 10-12 September 2008
10
4
10
2
3
10
10
4
Frequency (Hz)
International
Telecommunication
15
Union
Conclusions
The work highlighted
the challenges of
developing, using and
interpreting artificial
ears to predict the
acoustic performance
of mobile phones in
the ‘hand-held’
position
Even more apparent
when extending
measurement beyond
the typical narrowband frequencies
Lannion, France, 10-12 September 2008
The multiple statistical
analysis methods
described here give
more perspectives on
what the “average”
human impedance
response target is
Discussion ongoing
within ITU-T SG12 to
make use of these
results for current and
future
recommendations
International
Telecommunication
16
Union
Thank You
Acknowledgements
Nokia would like to thank the fellow
contributors to the measurements
presented herein, which included:
Brüel & Kjær (Denmark)
HEAD acoustics (Germany)
Motorola (USA)
Uniden (USA)
Lannion, France, 10-12 September 2008
International
Telecommunication
17
Union