SPPA 4030
Speech Science
Stephen M. Tasko Ph.D. CCC-SLP
Topic 1: The Speech Chain
Learning Objectives
• Outline the general sequence of
biological/physical events that occur from speech
formulation to speech perception.
• Describe the different types of information
content embedded within the speech signal.
• Know and describe the different branches of
physics and biology used to inform basic
mechanisms of speech production and
perception.
The Speech Chain (Denes & Pinson, 1993)
What information is embedded in the
speech signal?
•
•
•
•
•
Phonetic information
Affective information
Personal information
Transmittal information
Diagnostic Information
Branches of science employed to
understand speech communication
Physics
• Acoustics
• Aerodynamics
• Kinematics
• Dynamics
Biology
• Anatomy
–
–
–
–
Gross anatomy
Microscopic anatomy
Molecular biology
Neuroimaging
• Physiology
– Electrophysiology
Physical Quantities Review
An Independent Learning Activity
Learning Objectives
• Distinguish between basic and derived units
• Distinguish between scalar and vector
quantities
• Define a range of derived quantities with
special emphasis on displacement, velocity,
acceleration, force, pressure, intensity,
resistance and their physical relationship
Assignment 1
• See Assignments section of course website
• Due September 12, 2013
Topic 2: The Source-Filter Theory of
Speech Production: An Introduction
Learning Objectives
• Outline the key assumptions of the source filter
theory of speech production
• Distinguish between the source signal, filter
characteristics, and the output signal
• Use a range of examples to demonstrate
understanding of the source filter theory
• Distinguish to role that different vocal tract
structures play in speech sound generation and
speech sound filtering
Producing Speech
• The vocal tract can be
conceived as a set of
interconnecting tubes
and valves.
• Speech production is
achieved through the
systematic regulation of
air pressures and flows
within the vocal tract.
Source-Filter Theory of
Speech Production
• The sounds we hear as speech is the product
of a sound source that has undergone filtering
by the vocal tract
• source and the filter may be considered to be
independent of each other
Vocal tract is a tube that can vary its shape
From Titze (1994)
Source Filter Theory
Source
(Phonation)
Filter
(Resonator)
Speech
(What We Hear)
Input Spectrum
Frequency Response
Curve
Output Spectrum
Same Source, Different Filter
Different Source, Same Filters
White Noise
Different Source, Same Filters
(Human)
burp
Different Source, Same Filters
(Human)
snore
Different Source, Same Filters
(Human)
Lip buzz
Different Source, Same Filters
(Human)
?
Different Source, Same Filters
(Non-Human)
sheep
Different Source, Same Filters
(Non-Human)
accordion
Different Source, Same Filters
(Non-Human)
If it quacks like a duck…
Source Filter Theory Applied:
Alaryngeal Speech
Source-Filter Theory Applied:
Esophageal Insufflation Test
Source-Filter Theory Applied:
Tracheoesophageal (TE) Speech
Source Filter Theory Applied:
The Talkbox
Source-Filter Theory Applied:
The Talkbox
http://www.youtube.com/watch?v=YS3gAVNlceg
Topic 3: A Brief Review of
Physical Acoustics
Learning Objectives
• Outline the physical processes underlying simple harmonic motion
using the mass-spring model
• Describe the molecular basis of sound wave propagation
• Define the key characteristics of sinusoidal motion including
– Amplitude: instantaneous, peak, peak-to-peak, root-mean-square
(RMS), the decibel scale
– Frequency/period including units of measure
– Phase
– Wavelength
• Briefly describe the relation between the sine wave and uniform
circular motion
• Outline the relationship between the frequency and wavelength of
a sound wave
Spring Mass Model
• Mass (inertia)
– Newton’s first law of motion
– Opposition to
acceleration/deceleration
• Elasticity
– Opposition to displacement
– Rest position
– Recoil force
• Friction
http://phet.colorado.edu/en/simulation/mass-spring-lab
What is sound?
• It may be defined as the propagation of a
pressure wave in space and time.
• Sound must propagate through a medium
Sound-conducting media
• Medium is composed of
molecules
• Molecules have “wiggle room”
• Molecules exhibit random
motion
• Molecules can exert pressure
A
B
Model of air molecule vibration
(Time 1)
Air molecules sitting side by side
Rest positions
Model of air molecule vibration
(Time 2)
Model of air molecule vibration
(Time 3)
Model of air molecule vibration
(Time 4)
Model of air molecule vibration
(Time 5)
Model of air molecule vibration
a
Time
1
2
3
4
5
Distance
b
c
d
Wave action of molecular
motion
Time
1
2
3
4
5
Distance
Amplitude waveform
Position
Time
Amplitude waveform
Amplitude
Time
Where is pressure in this model?
Time
1
2
3
4
5
Pressure measuring device at
a specific location
Pressure waveform
Sound Pressure
+
0
Ambient Pressure
Time
Measuring Sound
•
•
•
•
Amplitude
Frequency
Phase
Wavelength
Measuring Sound: Signal Amplitude
Ways to measure it
• Instantaneous
• Peak
• Peak-to-peak
• Root mean square
(RMS)
• Decibel –see later
Sound Pressure
+
0
Time
Measuring Sound: Signal Amplitude
• Root mean square (RMS)
What units do we use to measure
signal amplitude?
• Pressure: Force/area
• Intensity = Power/area where
power=work/time & work=Force*distance
• Intensity is proportionate to Pressure2
Brief Review: The decibel scale
• decibel scale typically used to represent signal
amplitude
• Many common measurement scales are
absolute and linear
• However, the decibel scale is relative and
logarithmic
Absolute vs. relative measurement
• Relative measures are a ratio of a measure to
some reference
• Relative scales can be referenced to anything
you want.
• decibel scale doesn’t measure amplitude
(intensity or pressure) absolutely, but as a ratio
of some reference value.
Typical reference values
• Intensity
– 10-12 watts/m2
– Threshold for normal hearing at 1000 Hz
• Sound Pressure Level (SPL)
– 20 micropascals
However…
• You can reference intensity/pressure to
anything you want
For example,
• Post therapy to pre therapy
• Sick people to healthy people
• Sound A to sound B
Linear vs. logarithmic
• Linear scale: 1,2,3…
• For example, the difference between 2 and
4 is the same as the difference between 8
and 10.
• We say these are additive
Linear vs. logarithmic
• Logarithmic scales are multiplicative
• Recall from high school math and hearing science
10 = 101 = 10 x 1
100 = 102 = 10 x 10
1000= 103 = 10 x 10 x 10
0.1 = 10-1 = 1/10 x 1
Logarithmic scales use the exponents for the number scale
log1010 = 1
log10100 = 2
log 101000=3
log 100.1 = -1
Logarithmic Scale
• base doesn’t have to be 10
• In the natural sciences, the base is often 2.7…
or e
Logarithmic Scale
• Why use such a complicated scale?
– logarithmic scale squeezes a very wide range of
magnitudes into a relatively compact scale
– this is roughly how our hearing works in that a
logarithmic scales matches our perception of
loudness change
Combining the idea of logarithmic and
relative…
bel= log 10(Im/ Ir)
Im –measured intensity
Ir – reference intensity
A bel is pretty big, so we tend to use decibel where
deci is 1/10. So 10 decibels makes one bel
dBIL = 10log 10(Im/ Ir)
Intensity vs. Pressure
• Intensity is difficult to measure.
• Pressure is easy to measure – a microphone is
a pressure measuring device.
• Intensity is proportionate to Pressure2
Extending the formula to pressure
Using some logrithmic tricks, this translates our
equation for the decibel to
dBSPL= (2)(10)log 10(Pm/ Pr) = 20log 10(Pm/ Pr)
Measuring Sound: Frequency/period
Sound Pressure
+
0
-
Period (T)
Time
Period (T): duration
of a single cycle
Frequency (F):
rate that cycle
repeats itself (1/T)
Measuring Sound: Frequency/period
• Absolute measure
– Cycles-per-second: Hertz (Hz)
• Relative measure
– Octave (double or halving of frequency)
– Semitones (12 semitones = 1 octave)
Phase: Uniform Circular Motion
Initiating a sound waves that differ
only in phase
A force is applied to molecule at frequency f and time t
same force applied at frequency f at time t+a where a < the period of vibration
Spatial variation in pressure
wave
wavelength () is the
distance covering
adjacent high and low
pressure regions
Spatial variation in pressure wave
Time
1
2
3
4
5
Spatial variation in pressure
wave
Relation between frequency and
wavelength
=c/F where
: wavelength
F: is the frequency
c: is sound speed in medium (35,000 cm/sec)
Learning Objectives
• Draw and describe time-domain and frequency-domain
representation of sound
• Distinguish between simple and complex sound sounds with
regard to physical characteristics and graphical
representations
• Distinguish between periodic and aperiodic sounds with
specific emphasis on terms such as fundamental
frequency/period, harmonics, and overtones
• Distinguish between continuous and transient sounds
• Describe how waves sum, define Fourier's theorem and be
able to describe the basics of Fourier analysis
Graphic representation of
sound
• Time domain
– Called a waveform
– Amplitude plotted as a
function of time
• Frequency domain
– Called a spectrum
– Amplitude spectrum
• amplitude vs. frequency
– Phase spectrum
• phase vs. frequency
– May be measured using a
variety of “window” sizes
Same sound, different graphs
Time domain
Frequency domain
From Hillenbrand
Classification of sounds
• Number of frequency components
– Simple
– Complex
• Relationship of frequency components
– Periodic
– Aperiodic
• Duration
– Continuous
– Transient
Simple periodic sound
• Simple: one frequency component
• Periodic: repeating pattern
• Completely characterized by
– amplitude
– period (frequency)
– phase
• Other names: sinusoid, simple harmonic
motion, pure tone
Simple periodic sound: Graphic
appearance
From Hillenbrand
Complex periodic sounds
•
•
•
•
Complex: > one frequency component
Periodic: repeating pattern
Continuous
Frequencies components have a special relation
– Lowest frequency: fundamental frequency
• Symbol: fo
• Frequency component with longest period
– Higher frequency components: harmonics
• integer (whole number) multiples of the fo
Complex periodic sounds: Graphic
appearance
• Time domain:
– repeating pattern of pressure change
– within the cycle, things look complex
• Frequency domain:
– spectral peaks at evenly spaced frequency
intervals
• Auditory impression: sounds ‘musical’
Complex periodic sounds: Graphic
appearance
From Hillenbrand
Amplitude
Amplitude
Glottal Source
Time
Frequency
Amplitude vs. Phase
Spectrum
Amplitude spectrum:
different
Phase spectrum:
same
From Hillenbrand
Amplitude vs. Phase
Spectrum
Amplitude spectrum:
same
Phase spectrum:
different
From Hillenbrand
(Complex) Aperiodic sounds
• Complex: > one frequency component
• Aperiodic: Does not repeat itself
• Frequency components are not systematically
related
• May be
– Continuous
– Transient
Aperiodic sounds: Graphic
appearance
• Time domain:
– no repeating pattern of pressure change
• Frequency domain:
– the spectrum is dense
– No “picket fence”
• Auditory impression: sounds ‘noisy’
Aperiodic sounds: Graphic
appearance
From Hillenbrand
Analysis of complex waves
• Waves can be summed
• Complex waves are the sum of simple waves
• Fourier: French Mathematician:
– Any complex waveform may be formed by summing sinusoids of
various frequency, amplitude and phase
• Fourier Analysis
– Provides a unique (only one) solution for a given sound signal
– Is reflected in the amplitude and phase spectrum of the signal
– Reveals the building blocks of complex waves, which are sinusoids
Learning Objectives
• Draw and differentiate the waveform and the
waveform envelope
• Draw and differentiate the amplitude
spectrum, the phase spectrum and the
spectrum envelope
• Differential between short-term spectra and
long-term average spectra.
The “envelope” of a sound
wave
• Waveform envelope:
– imaginary smooth line that follows the peak of the
amplitude of a sound pressure waveform
• Spectrum envelope:
– Imaginary smooth line drawn on top of the
amplitude spectrum
Waveform envelope
From Hillenbrand
Waveform envelope
Time
Amplitude
Spectrum envelope
Frequency
Thought Question
Can an aperiodic and complex
periodic sound have identical
spectrum envelopes?
Amplitude Spectrum:
Window Size
• “short-term” vs. “long-term average”
amplitude spectrum
“Instantaneous” Amplitude
Spectra
(Long Term) Average Amplitude Spectrum
Learning Objectives
• Describe how the amplitude spectrum and the spectrogram
are related.
• Identify the axis units of the spectrogram.
• Provide some advantages of the spectrogram over the
amplitude spectrum.
• Distinguish between a wide band and narrow band
spectrogram and outline the different information each
provides.
• Distinguish between a harmonic and a formant on a
spectrogram.
• Be able to draw stylized (highly simplified) spectrograms
based on spectra and spectrum envelopes.
The Spectrogram
Building a spectrogram
F
A
Rotate
90 degrees
F
A
Building a spectrogram
F
Rotate it so that
The amplitude is
Coming out of the
page
F
A
Time
This is really narrow because it is a slice in time
Building a spectrogram
Frequency
Dark bands
= amplitude
Peaks
Time
Two main types of
spectrograms
• Narrow-band spectrograms
– Akin to amplitude spectrums “lined up”
– Frequency resolution is really sharp
• Wide-band spectrograms
– Akin to spectrum envelopes “lined up”
– Frequency resolution not so sharp
Wide vs. Narrow Band
Spectrograms
Highlights harmonic structure
Highlights spectrum envelope
Learning Objectives
• Define an acoustic filter
• Draw and label a frequency response curve
• Draw and differentiate different types of acoustic
filters
• Define terms such as cutoff frequency, center
frequency, roll off rate, gain, and bandwidth
• Define and draw a basic filter system and relate that
to the source-filter theory of speech production
What is an “Acoustic” Filter
• holds back (attenuates) certain sounds and lets other
sounds through - selective.
Why might we be interested in
filters?
• Human vocal tract acts like a frequency selective
acoustic filter
• Human auditory system behaves as a frequency
selective filter
• helps us understand how speech is produced and
perceived.
Kinds of frequency selective
filters
Low-pass filters
– Lets low frequencies “pass through” and attenuates high
frequencies
High-pass filters
– Lets high frequencies “pass through” and attenuates low
frequencies
Band-pass filters
– Lets a particular frequency range “pass through” and
attenuates other frequencies
Low Pass Filters
Gain
+
low
Frequency
high
High Pass Filters
Gain
+
low
Frequency
high
Band Pass Filter
Gain
+
low
Frequency
high
Frequency Response Curve (FRC)
Center frequency
+
3 dB
Gain
passband
lower cutoff
frequency
upper cutoff
frequency
low
high
Frequency
Operation of a filter on a
signal
NOTE:
Amplitude spectrum describes a sound
Frequency response curve describes a filter
Learning Objectives
• Define resonance, free and forced vibration
• Describe how the pendulum and spring mass models
can help explain resonance.
• Outline how mass and stiffness influences the resonant
frequency of a mass spring system.
• Outline how acoustic resonators behave like acoustic
filters.
• Calculate resonant frequencies of a uniform tube based
on its physical dimensions.
• Describe how the wavelength of the sound determines
the resonant frequency of tube.
Free vibration
• objects tend to vibrate at a characteristic or
resonant frequency (RF)
Forced vibration
• A vibrating system can force a nearby system
into vibration
• The efficiency with which this is accomplished
is related to the similarity in the resonant
frequency (RF) of the two systems
Forced vibration
• If the RF of the two systems are the same, the
amplitude of forced vibration will be large
• If the RF of the two systems are quite
different, the amplitude of forced vibration
will be small or nonexistent
Resonance refers to
• The tendency for an object to vibrate at a
particular frequency or frequencies.
• The ability of a vibrating system to force
another system into vibration.
Back to the mass spring
model
• Vibratory frequency of
the mass spring
determined by
– Mass
– Stiffness of the spring
http://phet.colorado.edu/en/simulation/mass-spring-lab
Acoustic Resonance
• Ideas from mechanical resonance also applies
to acoustic systems
• Acoustic chambers will transmit sound
frequencies with more or less efficiency,
depending upon the physical characteristics
• Therefore, they act as filters, passing through
(and even amplifying) some frequencies and
attentuating others.
Acoustic Resonance
• And since they act as filters, they have most of
the same features of a filter, even though we
might use different names.
• Center frequency is often termed the resonant
frequency.
• Frequency response curve often termed the
resonance curve.
Helmholtz Resonator
Actions of a Helmholtz
Resonator
Other Acoustic Resonators: Tube
Resonators
• Uniform tubes: Factors that influence resonance
– Length.
– Cross-sectional area along its length.
– Whether it is closed at either or both ends.
Stephen M. Tasko
Uniform tube, closed at one end
Uniform tube, closed at one end
Uniform tube closed at one end
First resonance or formant Higher resonant/formant
frequencies are odd
multiples of F1
F1 = c/4l
For example,
Where
• F1 = (c/4l )*1
c=speed of sound
(35,000 cm/sec)
• F2 = (c/4l )*3
l = length of the tube
• F3 = (c/4l )*5
males ~ 17.5 cm
females ~ 14 cm
Stephen M. Tasko
Comparing Helmholtz and
tube resonators
Resonator Features
Sharply tuned
Broadly tuned
Resonator Features
Gain
Frequency
An example of the resonance characteristics
of the human vocal tract