Figure%205.1.%20Measured%20synthetic%20fundamental%20period%20combines%20portions%20from%20two%20separately%20synthesized%20pulses;%20time%20between%20minima%20is%20measured:%20the%20last%20part%20of%20pulse%20T1,%20with%20length%20(1-a)T1,%20is%20combined%20with%20the%20first%20part%20of%20pulse%20T2,%20with%20length%20aT2.

1
Figure 5.1. Measured synthetic fundamental period
combines portions from two separately synthesized
pulses time between minima is measured the last
part of pulse T1, with length (1-a)T1, is
combined with the first part of pulse T2, with
length aT2. The resulting sum of two random
variables is subject to basic rules of error
propagation.
2
Figure 5.2 . Simulation demonstrating the effect
of fundamental peak position on standard
deviation of measured fundamental period length.
Circles show average simulated result using
100,000 random pulses. Line shows predicted
values.
3
Figure 5.3. Verification of NSR level in
synthesized voice. Aspiration noise (AN) levels
programmed predict measured levels within about 1
dB.
4
2
1.8
1.6
1.4
1.2
HFPV IN SYNTHETIC VOICE WITH ASPIRATION()
1
0.8
0.6
0.4
0.2
0
0
0.5
1
1.5
2
HFPV IN SYNTHETIC VOICE WITHOUT ASPRIATION ()
Figure 5.4. Effect of AN on measured HFPV in the
synthetic voice. Horizontal axis is programmed
HFPV level. Vertical axis is HFPV measured in
synthetic voice when the original measured level
of AN is added. AN seems to add about 0.2 to
the HFPV measurement for the levels found in the
voice set.
5
0
-5
-10
-15
NSR IN SYNTHEITIC VOICE WITH HFPV (dB)
-20
-25
-30
-30
-25
-20
-15
-10
-5
0
NSR IN SYNTHETIC VOICE WITHOUT HFPV (dB)
Figure 5.5 Effect of HFPV on measured NSR in the
synthetic voice. Horizontal axis is programmed
AN level. Vertical axis is measured NSR in the
synthetic voice when the original measured level
of HFPV is added. HFPV seems to add about a 4 dB
increment in NSR for the levels found in the
voice set. Dashed line shows best linear fit.
6
0
ORIGINAL VOICE
-5
PEARSON 0.51
-10
-15
SABS LISTENER-SET ASPIRATIONN NOISE (dB)
-20
-25
-30
-30
-25
-20
-15
-10
-5
0
MEASURED NSR (dB)
Figure 5.6. Mean of listener-set aspiration
noise in SABS experiments versus the measured
original NSR. The listeners set AN levels 5 - 10
dB below measured, indicating original NSR
includes more than AN effects. Dashed line shows
best linear fit.
7
0
TREMOR REMOVED
-5
PEARSON 0.71
-10
-15
SABS LISTENER-SET ASPIRATION NOISE (dB)
-20
-25
-30
-30
-25
-20
-15
-10
-5
0
MEASURED NSR (dB)
Figure 5.7. Mean of listener-set aspiration
noise in SABS experiments versus NSR of voices
with tremor removed. Agreement between SABS and
measured NSR improves over Fig. 5.6. Dashed line
shows best linear fit.
8
0
ALL AMFM REMOVED
-5
PEARSON 0.87
-10
SABS LISTENER-SET ASPIRATION NOISE (dB)
-15
-20
-25
-30
-30
-25
-20
-15
-10
-5
0
MEASURED NSR (dB)
Figure 5.8. Mean of listener-set aspiration
noise in SABS experiments versus NSR of voices
with all AM and FM removed. Agreement between
SABS and measured NSR improves further over Fig.
5.6. Dashed line shows best linear fit.
9
3
SABS HFPV
2.5
PEARSON 0.40
2
SABS LISTENER-SET HFPV ()
1.5
1
0.5
0
0
0.5
1
1.5
2
2.5
3
MEASURED HFPV ()
Figure 5.9. SABS comparison of user set HFPV
versus measured HFPV. Weak correlation indicates
listener difficulty perceiving HFPV at the levels
present in the voice set.