Heinz, J. M. and Lindblom, B. and. Lindqvist-Gauffin, J. journal: STL-QPSR volume: 8 .... J. Tobias and E. Schubert). (3) Lindblom, B. and Studdert-Kennedy, ...
Dept. for Speech, Music and Hearing
Quarterly Progress and Status Report
Patterns of residual masking for sounds with speech-like characteristics Heinz, J. M. and Lindblom, B. and Lindqvist-Gauffin, J.
journal: volume: number: year: pages:
STL-QPSR 8 2-3 1967 066-071
http://www.speech.kth.se/qpsr
SPEECH SYNTHESIZER
--, D/A
LPF
i
2
GATE 1
ATT ENUATOR A
AMP
+ MIXER
CHANNEL1 OUTPUT
O
,
LPF
GATE 2
AT TENUATOR A
EXPERIMENT CONTROL
Fig. 111-A- I .
Block diagram of inrtrumentation.
b
AMP
CHANNEL 2 \OUTPUT O
STL-QPSR 2-3/1967 Stimulus p a r a m e t e r s Fig. 111-A-2 i l l u s t r a t e s a t e s t i t e m typical of the present experiments and specifies the stimulus p a r a m e t e r s . The computer generates two sounds: a sinusoidal r a m p and a steady-state probe o r comparison tone.
The sweep tone whose amplitude i s A1 v a r i e s linearly between
i t s t e r m i n a l frequency values: fi and f f .
Its duration, T1, i s defined
i n t e r m s of the generated waveform r a t h e r than i n t e r m s of the timing of the on- and off-commands f o r the gates. off
T
Thus the g a t e s a r e turned
m s e c before the physical end of the signal.
The value of
T
is
chosen s o a s t o match the t i m e constant of t h e closing of the gates. To minimize clicks this constant, a s well a s that f o r the opening, was adjusted t o 5 m s e c throughout the following experiments.
The steady-
state tone follows the r a m p a f t e r an interval of T2 m s e c .
Its duration,
T3, i s defined i n analogy with T1.
In a pitch-matching t a s k the f r e -
quency of this tone, f t , can be controlled by the o b s e r v e r by means of the potentiometer connected to the computer. Alternatively, i n a n experiment on threshold determination he c a n s i m i l a r l y control the amplitude of i t , A2.
Finally, a latency period, T4, i s tagged on t o
the sequence of stimulus events.
During this period which was s e t a t
one second the stimuli cannot be repeated.
The end of this interval is
indicated t o the o b s e r v e r b y a lamp signal.
T h r e e o b s e r v e r s w e r e used i n the experiments to be reported. They all had n o r m a l audiog r a m s .
In preliminary experiments t h e i r
ability to d i s c r i m i n a t e pitch was found to be within 0.5 percent.
No
one had participated before a s a subject i n psychoacoustic experiments and no one was a w a r e of the purpose of the experiments.
During the
s e s s i o n s an o b s e r v e r was seated i n a quiet room listening monaurally to the stimuli over earphones and communicating with the computer by means of the lamp and push-buttons.
After the computer had recorded
a given response f r o m the subject i t would move on t o produce a following stimulus item. minutes.
This procedure would be repeated f o r 10 o r 15
The total duration of a complete s e s s i o n exclusive of b r e a k s
never amountcd to m o r e than 30-45 minutes.
The sound-pressure level of the stimuli was 85 dB rcl. 0.0002 dynes/cm 2 i n a l l the experiments.
DIRECTION OF RAMP: ASCENDING I
1
DESCENDING I I I
kHz
f
I
-
iI
-
I
-
I
I
-
I I I
-
I
I I
EXTENT OF RAMP: 1.0 - 1.5 kHz
I I I
20
I
50
I 1
I
20
I
1
50 msec
DURATION OF RAMP
Fig. 111-A-3.
Results of matching the pitch of a steady tone to that of a sliding tone. The median frequency of the pitch matches made by three observers are shown together with the interquartile ranges for four etimulue conditions.
STL-QPSR 2-3/1967
Table 111-A-I.
Individual results of pitch matches f o r t h r e e observers.
Table entries r e f e r t o the median
frequencies of five observations.
Variational
ranges i n parentheses.
f
Direction and duration ( m s e c ) of ramp:
20 TE
RE
i
JS
.
Descending
Ascending
Observer I
50
20
1350 (30)
1370 (60)
1175(30)
1550 (405) 1465 (1 10)
1510 (95) 1430 (305)
1205 (235) 1005 (60)
-4
50
/
1150(50) 1100 (110) 990 (90)
1 t
STL-QPSR 2-3/1967
70.
Discussion: P e r c e p t u a l extrapolation The notion of perceptual extrapolation has recently attracted X2X3). F o r instance, i t was put f o r some attention i n the l i t e r a t u r e ward by Brady et a1 t o account f o r the finding that the p r e f e r r e d matching frequency tended t o be c l o s e r to the t e r m i n a l value of the formant r a m p when the duration of the r a m p was altered f r o m 50 t o 20 m s e c and consequently the r a t e of formant frequency change was increased. On the assumption of a constant integration t i m e of the e a r i t was a r gued that a value c l o s e r to the average value of the r a m p could have Moreover, under c e r t a i n conditions the frequency of
been expected.
the match was often beyond the final resonant frequency of the stimulus. In conclusion t h e s e investigators suggest that "there may be s o m e overshoot o r extrapolation i n the processing of brief stimuli c h a r a c t erized by rapidly changing spectra". In this context t h r e e experiments w e r e performed t o t e s t , i n a p r e liminary way, the r e a l i s m of the following question: does the poststimulatory threshold pattern associated with a sweep-tone extend i n frequency beyond the pattern corresponding to a tone stationary a t the ramp terminal? Residual masking patterns of r a m p stimuli The patterns of residual masking f o r two sweep-tone stimuli and a 1500 Hz tone have s c f a r been determined f o r one of the subjects.
The
r a m p s varied f r o m 1000 t o 1500 Hz and downwards f r o m 2000 to 1500 Hz.
The interstimulus delay, T2, was 5 m s e c (cf."Section on Stim-
ulus p a r a m e t e r s t 1 ) . The duration of the t e s t stimuli was 50 m s e c while that of the probe The o b s e r v e r could control the amplitude
was 20 m s e c i n all c a s e s . of the probe.
T e n frequencies w e r e selected f o r the probe tone i n the
range of 750 to 3000 Hz.
E a c h probe frequency occurred five t i m e s .
In each t e s t the computer presented a randomized sequence of 50 stimulus pairs.
The subject was instructed to s e t the amplitude of the
probe tone a s c l o s e a s he could to the threshold.
Figs. 111-A-4 and
111-A-5 demonstrate the r e s u l t s of t h e s e measurements.
These plots
show the mean threshold locations relative to the 85 dB t e s t stimulus a s a function of frequency. i n both graphs.
Filled c i r c l e s r e p r e s e n t the 1500 Hz c a s e
In Fig. 111-A-4 the residual masking of the 1000-1500
Hz sweep has been superimposed; i n Fig. 111-A-5 that of the 2000-1500
MEDIAN FREQUENCY OF MATCH
FREQUENCY
F i g . 111-A-4.
Residual masking p a t t e r n s a s s o c i a t e d with 1500 Hz tone and a s i n u s oidal r a m p varying f r o m 1000 t o 1500 Hz both of 50 m s e c duration. P e l a y = 5 m s e c . Stimulus intensity 85 dB r e l . 0.0002 dynes/crn2.
MEDIAN FREQUENCY OF MATCH
FREQUENCY
Fig. 111-A-5.
Residual masking patterns associated with 1500 Hz tone and a qinus. oidal ramp varying f r o m 2000 to 1500 H e both of 50 m s e c durati Delay = 5 m s e c . Stimulus intensity 85 dB rel. 0.0002 dyner/cm'=P
.
STL-QPSR 2-3/1967
Hz ramp i s shown.
The following observations can be made:
1. The patterns for the sweep tones resemble that of the 1500 Hz tone and do not show uniform masking over the frequency range of the sweep. 2.
These patterns display maxima of masking a t frequencies that approach the median frequencies of the pitch matches (6 observations i n each case).
3.
In neither c a s e does the sweep-tone pattern exhibit g r e a t e r masking i n the frequency region beyond the terminal value of the ramp.
References: (1) Brady, P. T . , House, A. S. , and Stevens, K. N. : I1Perception of Sounds Characterized by a Rapidly Changing Resonant >?requencyl', J. Acoust. Soc. Am., 33 (1961), pp. 1357-1362.
-
(2) Stevens, K.N. and House, A.S.: I1Speech Perception", a chapter t o be publ. in Foundations of Modern Auditory Theory (eds. J. Tobias and E. Schubert).
(3) Lindblom, B. and Studdert-Kennedy, M. : "On the R61e of Formant Transitions in Vowel Recognition", to be publ. i n J. Acoust. Soc. Am. , fall 1967.
