Room Acoustics: implications for
speech reception and perception
by hearing aid and cochlear
implant users
Arthur Boothroyd, Ph.D.
Distinguished Professor Emeritus, City University of New York
Scholar in Residence, San Diego State University
Visiting Scientist, House Ear Institute
2003
Part A
Room acoustics
and speech
audibility
Critical Factors
1. Speech spectrum
2. Noise
3. Distance
4. Early reverberation
(early reflections)
5. Late reverberation
1.
The spectral
distribution of useful
information in the
original speech signal
Speech spectrum at 1 foot
70 dBSPL
1 foot
Speech spectrum at 1 foot
Long-term level, broad-band rms
Long-term level, 1/3 octave rms
Measured peak level, 1/3 octave rms
Speech level in dBSPL
Idealized short-term (50 ms) range
80
70
60
50
40
30
20 Relative
10 importance
0
4%
125
250 15%
500 30%
1000 35%
2000 15%
4000 1%
8000
Frequency in Hz
Speech spectrum at 1 foot
Long-term level, broad-band rms
Long-term level, 1/3 octave rms
Measured peak level, 1/3 octave rms
Speech level in dBSPL
Idealized short-term (50 ms) range
80
70
60
50
40
30
20
10
0
4%
125
250 15%
500 30%
1000 35%
2000 15%
4000 1%
8000
65%
95%
Frequency in Hz
Speech spectrum at 1 foot
Distance 1 feet
Average level
= 72 dBSPL
30 dB
Range of most
useful
information = 30
dB
(from about 40
to about 70
dBHL)
2.
Background Noise
Percent Audibility
Speech Audibility and the
signal-to-noise ratio
100
80
30dB
60
40
20
0
5 10 15
-20 -15 -10 -5 0
Signal-to-noise ratio in dB
20
Speech
Speech
Speech
level
increasing
30dB
Speech
Speech
Noise
The spectral distribution of useful
information in the original speech signal
• Most important range is from around 700 to around
3000 Hz - contains 65% of the useful information
• Extending down to 350 Hz and up to 6000 Hz adds
another 30% (total = 95%).
• Useful information covers a range of 30 dB
• Signal-toSpeech
Phoneme
noise ratio
Audibility
Recognition
-15 dB
0%
0%
0 dB
50%
85%
+15 dB
100%
98%
3.
The distance between
listener and talker
The distance between listener and
talker
• The level of the direct speech signal falls
by 6 dB for every doubling of distance
48 dBSPL
60dBSPL54 dBSPL
4 ft
8 ft
16 ft
42 dBSPL
32 ft
Direct speech at 32 ft
= 42 dBSPL
Direct speech at 4 ft
= 60 dBSPL
Direct speech at 1 ft
= 72 dBSPL
4.
The early components of
reverberation (early
reflections)
Early reverberation
• Reverberation is the persistence of sound in
an enclosed space because of multiple,
repeated reflections from the boundaries
The reverberation effect
Direct sound
Reverberation
Enclosed
space
Free field
The reverberation effect
Direct sound
Reverberation
The
shower
analogy
Early reverberation
• - the persistence of sound in an enclosed
space because of multiple, repeated
reflections from the boundaries
• Reverberation time (RT60) is the time taken
for the sound level to drop by 60 dB after the
source is turned off
80
Reverberation time (RT60)
60
Amplitude
40
in
dBSPL 20
0
80 0
500 msec
100
60
300
400
500
Time in milliseconds
Amplitude
40
in
dBSPL 20
0
200
0
100
200
300
400
Time in milliseconds
60 dB
500
Early reverberation
• Reverberation is the persistence of sound in
an enclosed space because of multiple,
repeated reflections from the boundaries
• Reverberation time (RT60) is the time taken
for the sound level to drop by 60 dB after the
source is turned off
• The early components of reverberation are
those reflections that arrive soon enough to
be integrated with the direct sound, and with
each other, so as to enhance perception
(less than 1/20 of a second)
Reverberation is not all bad - early reflections can increase
loudness without reducing clarity
Early reverberation
• Reverberation is the persistence of sound in an
enclosed space because of multiple, repeated
reflections from the boundaries
• Reverberation time (RT60) is the time taken for the
sound level to drop by 60 dB after the source is
turned off
• The early components of reverberation are those
reflections that arrive soon enough to be integrated
with the direct sound, and with each other, so as to
enhance perception
• The early components of reverberation increase
the level of speech at a distance
The early components of reverberation
Increase the level of speech at a distance
Critical
distance
Speech level in dBSPL
80
Direct speech signal
Early reverberation
Combined signal
70
60
50
Direct speech
negligible
40
30
20
0
5
10
15
20
25
Distance from talker in feet
30
35
Early reverberation
• Reverberation is the persistence of sound in an
enclosed space because of multiple, repeated
reflections from the boundaries
• Reverberation time (RT60) is the time taken for the
sound level to drop by 60 dB after the source is
turned off
• The early components of reverberation are those
reflections that arrive soon enough to be integrated
with the direct sound, and with each other, so as to
enhance perception
• The early components of reverberation increase
the level of speech at a distance
5.
The late components of
reverberation
The late components of
reverberation
• They arrive too late to be integrated with the
direct signal or the early components (more
than 1/10 of a second)
• If their level is still high enough, they interfere
with the current sound by both physical and
perceptual masking
Text analogy
Late
The
Thefollowing
followingisisaalist
listofofFarmer’s
Farmer’s
markets
marketstotobe
beheld
heldininthe
thesurrounding
surrounding
areas
areas
Early
The
The following
following is
is aa list
list of
of Farmer’s
Farmer’s
markets
markets to
to be
be held
held in
in the
the surrounding
surrounding
areas
areas
Severe
reverberation
Where can I
suit
get my
cleaned?
8
Input
to
listener
Frequency in kHz
Output
from
talker
6
4
2
0
8
6
4
2
0
0.0
0.2
0.4
0.6
0.8
1.0
Time in seconds
1.2
1.4
1.6
1.8
The late components of
reverberation
• They arrive too late to be integrated with the
direct signal or the early components (more
than 1/10 of a second)
• If their level is still high enough, they interfere
with the current sound by both physical and
perceptual masking
• Effectively, the reverberant speech signal
generates its own masking
The late components of
reverberation
Speech
signal
30 dB
Self
masking
The late components of
reverberation
• They arrive too late to be integrated with the
direct signal or the early components (more
than 1/10 of a second)
• If their level is still high enough, they interfere
with the current sound by both physical and
perceptual masking
• Effectively, the reverberant speech signal
generates its own masking
• The effective signal-to-noise ratio depends on
the reverberation time
The late components of
reverberation
0.25
Reverberation time in seconds
0.5
1.0
2.0
4.0
8.0
30dB
12
7
2
-3
Effective signal-to-noise ratio in dB
-8
-13
90
73
Percent Audibility
23
7
98
95
91
81
62
Phoneme recognition in CVC words in %
24
57
40
Part B
The
Speech Audibility
Index
(SAI)
Speech Audibility Index
(SAI)
Combines effects of:
•Direct Speech Signal
&
Effective signal
&
Effective noise
•Early Reverberation
•Noise
•Late Reverberation
Effective s/n ratio
Speech Audibility Index (SAI)
• Speech Audibility Index is the proportion of
the combined direct speech signal and
early reverberation that exceeds the
combined noise and late reverberation.
• It rises from 0 (no useful speech audible)
to 100% (all useful speech audible) as the
effective signal to noise ratio rises from
-15 dB to +15 dB.
Part C
Effective signal-tonoise ratio,
Speech Audibility
Index and Speech
Perception
Speech Perception
1.Phonemes (vowels
and consonants)
2.Words in isolation
3.Words in sentences
Speech Audibility Index
0 20 40 60 80 100
100
1.0
80
0.8
60
0.6
40
0.4
y = (1 - .0054((x+15)/30))1.47
20
0.2
0
0.0
-20
-10
0
10
20
Signal-to-noise ratio in dB (x)
Phoneme recognition
probability (y)
Phoneme recognition
ptobability in %
Phoneme recognition in CVC
words
CVC words in isolation
Probability of recognizing
a phoneme = p
C1
V
w=
j
p
C2
j = 3 for unfamiliar
words
Probability of recognizing
whole word = w
j = 2 for familiar
words
Speech Audibility Index
0 20 40 60 80 100
100
1.0
80
0.8
13%pts.
%points
60
0.6
13
40
0.4
FamiliarFamiliar
words
Unfamiliar
words
20
0.2
Unfamiliar
22dB
dB
0
0.0
-20
-10
0
10
20
Effective signal-to-noise ratio in dB
Word recognition
probability
Word recognition
probability in %
CVC words in isolation
Words in sentences
Probability of recognizing a
word in isolation = w
s = 1 – (1-w)k
Lamb
Would you like Lamb for dinner
Probability of recognizing a
word in a sentence = s
k < 2 for complex
sentences
k > 5 for simple
sentences
Words in sentences
Speech Audibility Index in %
0 20 40 60 80 100
100
1.0
11.5
dB
11.5dB
80
0.8
Simple sentences
60 95%
Familiar words 0.6
Simple sentences,
40
0.4
Complex
sentences
familiar words
Complex sentences,
38 %
20
38%
Unfamiliar
words 0.2
unfamiliar words
0
0.0
-20
-10
0
10
20
Signal-to-noise ratio in dB
Word recognition
probability
Word recognition
probability in %
Normal
Hearing
Normal hearing
Words in sentences
Aided 50 dB Sensorineural Loss
Speech Audibility Index in %
0 20 40 60 80 100
Word recognition
probability
Word recognition
probability in %
Normal
100
1.0
6.5
Simple sentences
80
0.8
dB
Familiar words
60 95%
0.6
6.5 dB
50 dB unaided
40
0.4
Complex sentences
hearing loss
36
%
Unfamiliar words
20
0.2
36% (plus amplification)
0
0.0
-20
-10
0
10
20
Signal-to-noise ratio in dB
Speech Perception
• Speech Audibility Index  phoneme
recognition  isolated word recognition 
sentence perception
• Relationship between SAI and sentence
perception depends on word familiarity,
sentence complexity, cochlear pathology,
listener knowledge, listener skills
• But optimization of Speech Audibility Index
is a crucial first step
Optimizing SAI for hearing aid
and cochlear implant users
•
•
•
•
•
•
Reduce Background noise
Reduce reverberation time (- but)
Enhance early reflections (- but)
Use directional microphones (- but)
Use remote (wireless) microphones (- but)
Use beam-forming (super-directional)
microphones (- but)
• Use Sound-Field amplification (- but)
The
Sound Field
solution
Sound-field amplification
FM link
Amplifier
Loudspeakers
Increasing volume
DOES increase
signal-to-noise
ratio
Increasing volume DOES increase
signal-to-noise ratio
dBSPL
90
80
70
Direct sound
60
50
40
NOISE
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume DOES increase
signal-to-noise ratio
dBSPL
90
80
Direct sound
70
60
50
40
NOISE
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume DOES increase
signal-to-noise ratio
dBSPL
90
80
Direct sound
70
60
50
40
NOISE
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume DOES increase
signal-to-noise ratio
dBSPL
90
Direct sound
80
70
60
50
40
NOISE
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume
does NOT increase
signal-to-reverberation
ratio
Increasing volume does NOT increase
signal-to-reverberation ratio
dBSPL
90
80
70
Early reflections
60
50
40
Direct sound
Late reflections
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume does NOT increase
signal-to-reverberation ratio
dBSPL
90
80
Early reflections
70
60
50
Direct sound
40
Late reflections
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume does NOT increase
signal-to-reverberation ratio
dBSPL
90
80
Early reflections
70
60
50
Direct sound
Late reflections
40
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Increasing volume does NOT increase
signal-to-reverberation ratio
dBSPL
90
Early reflections
80
70
60
50
Direct sound
Late reflections
40
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Improving proximity
DOES increase
signal-to-reverberation
ratio
Improving proximity DOES increase
signal-to-reverberation ratio
dBSPL
90
80
70
Early reflections
60
50
40
Direct sound
Late reflections
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Phoneme recognition in CVC words by 14 child implant users.
(data courtesy of Frank Iglehart of the Clarke School for the Deaf)
(Bars show mean with
Standard error)
Improving directionality
(Q) DOES increase
Signal to reverberation
ratio
Low Directionality (Q)
dBSPL
90
80
70
Early reflections
60
50
40
Direct sound
Late reflections
30
20
10
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
High directionality (Q)
dBSPL
90
80
Early reflections
70
60
50
40
30
10
Direct sound
Late reflections
20
0
2
4
6
8
10
12
14 16
18 20
22
24
26 28
30
32
34
36 38
Distance from loudspeaker in feet
Amplifier
40
Summary re Sound-Field
15 to 20 dB increase of s/noise and s/reverb at mic.
Redistributed with possible increase of overall level.
Benefit to listener reduced by:
Distance to nearest loudspeaker (6 dB rule).
Room reverberation (increases with additional sources).
Room noise.
Increasing system gain increases signal/noise,
BUT NOT signal/reverberation.
Improving proximity increases signal/noise,
AND signal/reverberation.
Increasing loudspeaker directionality increases s/reverb.
If primary problem is noise, place speakers high for coverage.
If primary problem is reverb., place speakers low for proximity.
(for individuals, desk-top system may be the best)
Priority should always be to minimize reverberation.
Sound-field system can then address distance and noise.
Directional speakers are to be preferred.
Measuring
Reverberation
Times
Estimating
Reverberation
Times
Simulating
Sound-Field
Installation
Demonstrations
• Measuring reverberation time
• Simulating the effects of room acoustics
and Sound-Field amplification (SFWIZ –
downloadable free of charge from either
www.arthurboothroyd.com or
www.phonicear.com
Additional reading
Boothroyd A (2003). Room Acoustics and Speech Perception.
Seminars in Hearing (in press). Draft available on
www.arthurboothroyd.com