development of improved noise metrics and auditory risk assessment procedure
DESCRIPTION
Development of Improved Noise Metrics and Auditory Risk Assessment Procedure. June 22, 2009 Won Joon Song and Jay Kim Mechanical Engineering Department University of Cincinnati. Contents. Description of noise metric Correlation study between the noise metric and PTS data - PowerPoint PPT PresentationTRANSCRIPT
Development of Improved Noise Metrics and Auditory Risk Assessment Procedure
June 22, 2009
Won Joon Song and Jay KimMechanical Engineering Department
University of Cincinnati
Contents
• Description of noise metric• Correlation study between the noise metric
and PTS data• Methodology to determine spectral NIHL
threshold SPL• EARM curve and its implementation
strategy to NIHL research
Noise group data
• Provided by collaborators in SUNY Plattsburgh• 23 noise groups
– 18 noises of 100 dBA overall SPL– Additional 5 noises
• 3 noises of 90 dBA overall SPL: G-47, G-48 and G-56• 2 noises of 95 dBA overall SPL: G-57 and G-58
• Complex or Gaussian noise– 20 complex noises– 3 Gaussian noises: G-61, G-47 and G-57
Noise exposure data• Provided by collaborators in SUNY Plattsburgh• Continuous 5-day exposure to 9-16 chinchilla subjects
(23 groups, 275 chinchillas in total) • Obtained at 6 frequency points of 0.5, 1.0, 2.0, 4.0,
8.0 and 16.0 kHz– Auditory evoked potential (AEP) measurement
• TTS (dB): right after exposure• PTS (dB): after 30 days of recovery
– Mechanical damage• OHC loss (%)• IHC loss (%)
Basic designs of noise metric
• 6 basic designs reflecting time-frequency characteristics of the noise• 14 metrics derived from basic designs
Type 1: Equivalent SPL
Type 2: Kurtosis
Type 3: Maximum SPL
Type 4: Dynamic SPL
Type 5: Modified equivalent SPL
Type 6: Normalized weighted exposure time
1/3 octave pressure time
history
1/3 octave pressure time-history: Implementation of AWT
0 1 2 3-100
-50
0
50
100
Time (sec.)
Sou
nd P
ress
ure
(Pa)
Pressure time-history
1/3 octave pressure time-history
• A special version of analytic wavelet transform developed at UC
• Signal analyzer to obtain a set of 1/3 octave time histories from a single time history
• Useful tool to study spectral noise-to-hearing loss relationship
Noise metric calculation procedure
Pressure time-history
Spectral noise metric
1/3 octave time-history @ 6 frequency points
0.5 kHz 1.0 kHz 2.0 kHz
16.0 kHz8.0 kHz4.0 kHz
T-F decomposition by AWT
Noise metric description
2
0
1( ) ( , )T
eqP P t dtT
2
10 2
( )( ) 10log eq
eqref
PL
P
Type 1: Equivalent SPL
Time average of 1/3 octave sound pressure
Type 5: Modified equivalent SPL
1
0
1( , ) ( , )T
em thP P t P dtT
2
1
( ) ( , )N
eq ii
tP P tT
1
1
( , ) ( , )N
em i thi
tP P t PT
Time average of 1/3 octave sound pressure above threshold
10
( )( , ) 10log em
emref
PL
P
( , ) if ( , )( , )
0 if ( , )th th
thth
P t P P t PP t P
P t P
Noise metric description
4
144 2
2
1
1 ( , ) ( )( )
( )( ) 1 ( , ) ( )
N
i mi
N
i mi
P t PN
P t PN
Type 2: Spectral kurtosis
Statistical quantity representing impulsiveness of a signal
0 0.5 1 1.5 2 2.5 3-100
-50
0
50
100
Time (sec.)
Sou
nd P
ress
ure
(Pa)
β =79.81
0 0.5 1 1.5 2 2.5 3-100
-50
0
50
100
Time (sec.)
Sou
nd P
ress
ure
(Pa)
β =2.99 ≈ 3
G-63: Complex type
G-61: Gaussian type
Noise metric description
295%
max 10 2
( )( ) 10log
ref
PL
P
Type 3: Max. SPL
Top 95% value of the 1/3 octave SPL distribution histogram
Type 4: Dynamic SPL
Weight to the dynamic fluctuation of 1/3 octave SPL
max( ) ( ) ( ) ( )d m mL L K L L
Type 6: Normalized weighted exposure time
Application of ‘3 dB exchange rule’ to each sampling time interval
( , ) 3, ( ) 2 i thL t L
w i it t
,1
( )( )
N
w ii
w
tT
T
Box plot of PTS data
Noise vs. frequency correlationNoise correlation: frequency by frequency
Frequency correlation: noise by noise
Noise metric surface
PTS surface (median)
Example of noise metric vs. median PTS: Complex noise (G-63)
103
104
0
10
20
30
40
50
PTS
(dB
)
103
104
70
80
90
100
Leq
(dB
)
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
0
500
1000
1500
Kur
tosi
s
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
40
60
80
100
120
Lmax
(dB
)
Frequency (Hz)
103
104
0
10
20
30
40
50
PTS
(dB
)
103
104
40
60
80
100
120
Ld (d
B)
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
60
70
80
90
100
Lem
(dB
)Frequency (Hz)
103
104
0
10
20
30
40
50
103
104
0
10
20
30
40
50
Tw
Frequency (Hz)
Spectral kurtosisMedian PTS
Example of noise metric vs. median PTS: Gaussian noise (G-61)
103
104
0
10
20
30
40
50
PTS
(dB
)
103
104
80
85
90
95
Leq
(dB
)
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
3.2
3.25
3.3
Kur
tosi
s
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
85
90
95
100
Lmax
(dB
)
Frequency (Hz)
103
104
0
10
20
30
40
50
PTS
(dB
)
103
104
90
95
100
105
Ld (d
B)
Frequency (Hz)10
310
40
10
20
30
40
50
103
104
70
75
80
85
90
95
Lem
(dB
)Frequency (Hz)
103
104
0
10
20
30
40
50
103
104
0
5
10
15
20
25
Tw
Frequency (Hz)
Spectral kurtosisMedian PTS
Correlation study with 18 noises
Correlation scale:
Noise correlationFrequency correlation
emL
eqL
wT
maxL
dL( 2,3, 4,5,10)K
( 2,3,4,5,10)
Correlation study with 23 noisesNoise correlationFrequency correlation
emL
eqL
wT
maxL
dL( 2,3, 4,5,10)K
( 2,3,4,5,10)
Correlation scale:
NIHL threshold SPL determined by statistical process
Scatter plot of Leq-PTS (1.0 kHz)
Upper prediction bound
Lower prediction bound
Linear regression line
0 dB
PTS
C.I.
Lower prediction bound
Threshold SPL (1.0 kHz)
Threshold SPL curve
NIHL threshold SPL curve construction
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)Le
q (d
B)
Simultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous observation bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Nonsimultaneous functional bounds
0 10 20 30 40 5060
70
80
90
100
110
120
PTS (dB)
Leq
(dB
)
Simultaneous functional bounds
0.5 kHz
1.0 kHz 2.0 kHz
8.0 kHz
16.0 kHz
4.0 kHz
L.B. (PTS 0dB)( ) ( )th eqL L
Improved noise metric with variable threshold
10
( )( , ) 10 log em
emref
PL
P
1
0
1( , ) ( , ) ( )T
em thP P t P dtT
Type 5a: Modified equivalent SPL with variable threshold
Time average of 1/3 octave sound pressure time history above frequency-dependent thresholds
1
1
( , ) ( , ) ( )N
em i thi
tP P t PT
L.B. (PTS 0dB)( ) ( )th eqL L
T-F decomposition
Variation of the equivalent sound pressure
2
0
1( ) ( , )T
eqP P t dtT
1
0
1( , ) ( , )T
em thP P t P dtT
1
0
1( , ) ( , ) ( )T
em thP P t P dtT
2
0
1 ( )T
eqP P t dtT
Constant threshold
Variable threshold
Type 5a
Type 5
Type 1Conventional
Time averaging
Constant vs. variable threshold
Constant threshold
Variable threshold
Equal auditory risk metric (EARM) curve construction
103
104
65
70
75
80
85
90
Frequency (Hz)
Lem
(dB
)
Nonsimultaneous observation bounds
103
104
60
65
70
75
80
85
Frequency (Hz)
Lem
(dB
)
Simultaneous observation bounds
103
104
75
80
85
90
95
100
Frequency (Hz)
Lem
(dB
)
Nonsimultaneous functional bounds
103
104
75
80
85
90
95
Frequency (Hz)
Lem
(dB
)
Simultaneous functional bounds
Scatter plot of Lem-PTS (1.0 kHz)
Upper prediction bound
Lower prediction
bound
Linear regression line
C.I.
EARM curvesLower prediction bound
Interpretation of EARM curves
103
104
65
70
75
80
85
90
Frequency (Hz)
Lem
(dB
)
Nonsimultaneous observation bounds
103
104
60
65
70
75
80
85
Frequency (Hz)
Lem
(dB
)
Simultaneous observation bounds
103
104
75
80
85
90
95
100
Frequency (Hz)
Lem
(dB
)
Nonsimultaneous functional bounds
103
104
75
80
85
90
95
Frequency (Hz)
Lem
(dB
)Simultaneous functional bounds
Iso-PTS curves(10,20,30,40,50dB)
Slow development of PTS
Fast development of PTS
PTS-free zoneNIHL threshold metric curve (0 dB PTS curve)
Application of EARM curve to NIHL study: Noise reduction level
103
104
60
70
80
90
100
Frequency (Hz)
Lem
(dB
)
G247
103
104
0
50
100
150
200
250
300
Frequency (Hz)
PTS
(dB
): P
redi
cted
103
104
0
10
20
NR
L (d
B):
Rec
omm
ende
d
Frequency (Hz)
0 50 100 150
0
50
100
150
PTS (dB): Measured
PTS
(dB
): P
redi
cted
NIHL threshold metric curve (0 dB PTS curve)
103
104
60
70
80
90
100
Frequency (Hz)
Lem
(dB
)
G247
103
104
0
50
100
150
200
250
300
Frequency (Hz)
PTS
(dB
): P
redi
cted
0 50 100 150
0
50
100
150
PTS (dB): Measured
PTS
(dB
): P
redi
cted
103
104
0
10
20
NR
L (d
B):
Rec
omm
ende
d
Frequency (Hz)
Noise metric curve (G-47)
Recommended spectral noise reduction level to protect the auditory system
Noise reduction level required at 8.0 kHz
No noise reduction required at 2.0 kHz
Application of EARM curve to NIHL study: PTS prediction
103
104
60
70
80
90
100
Frequency (Hz)
Lem
(dB
)
G247
103
104
0
50
100
150
200
250
300
Frequency (Hz)
PTS
(dB
): P
redi
cted
103
104
0
10
20
NR
L (d
B):
Rec
omm
ende
d
Frequency (Hz)10
310
40
20
40
PTS
(dB
): M
ean
0 50 100 150
0
50
100
150
PTS (dB): Measured
PTS
(dB
): P
redi
cted
Noise metric curve (G-47)
PTS at 4kHz (estimated by interpolation)
0 dB PTS at 2.0 kHz
Validity checking of EARM curve prediction
103
104
60
70
80
90
100
Frequency (Hz)
Lem
(dB
)
G247
103
104
0
50
100
150
200
250
300
Frequency (Hz)
PTS
(dB
): P
redi
cted
103
104
0
10
20
NR
L (d
B):
Rec
omm
ende
d
Frequency (Hz)10
310
40
20
40
PTS
(dB
): M
ean
0 50 100 150
0
50
100
150
PTS (dB): Measured
PTS
(dB
): P
redi
cted
Properl
y-esti
mated
103
104
60
70
80
90
100
Frequency (Hz)
Lem
(dB
)
G247
103
104
0
50
100
150
200
250
300
Frequency (Hz)
PTS
(dB
): P
redi
cted
103
104
0
10
20
NR
L (d
B):
Rec
omm
ende
d
Frequency (Hz)10
310
40
20
40
PTS
(dB
): M
ean
0 50 100 150
0
50
100
150
PTS (dB): Measured
PTS
(dB
): P
redi
cted
Predicted PTS at 8.0 kHz
Measured PTS at 8.0 kHz
Overestimation at 0.5, 8.0, and 16.0 kHz
Under-es
timate
dOve
r-esti
mated
Acceptable PTS prediction band
Questions?