Comd 5070 Speech Science Week 3 Of 5 Exam 1

Front 1

What is the Fourier Transform?

Back 1

when we use the techniques that Fourier developed

Front 2

What does a Fourier Transform do?

Back 2

creates a spectrum from a time domain waveform - shows what the individual components are.

Example - prism/light
cake/ingredients

Front 3

Time Domain Waveform?
x and y axis?

Back 3

is time domain data, you see it when you have a microphone recording on a screen in front of you.
x axis-time
y axis-amplitude

Front 4

What do you get when you subject this time domain waveform to a Fourier Transform?

Back 4

a Frequency Domain Display - spectrum, slice of time/snapshot

Front 5

The Fourier Transform is a way from going to one view of a signal to another.
From the ____ Domain to the _____ Domain

Back 5

Time, Frequency

Front 6

Time domain data (2)

Back 6

a waveform that represents sound directly
air pressure changes over time

Front 7

Frequency domain data (spectrum display)

Back 7

line spectrum show the frequency components of a periodic sound

-we get it from the time domain and doing a Fourier transform on it to give us the spectrum display)

Front 8

Great benefit of Frequency Domain Display

Back 8

show us what individual components are and also the relative proportions of each

Front 9

What is a harmonic?

Back 9

the harmonics are exact multiples of that fundamental frequency
-there is no energy between the individual harmonic components that make up the entire spectrum

Front 10

that the human voice is nearly _____

Back 10

periodic

Front 11

It has a fundamental and then harmonics which are multiples of that fundamental. And just as you see the upper harmonic components getting shorter in height as the frequency gets higher, the same is true of the _____ ______. The ____ harmonics get progressively _____as you go up in frequency.

Back 11

Human Voice
upper
weaker

Front 12

PERIODIC SIGNAL SPECTRUM (4)

Back 12

-frequency domain description of the signal
-has harmonics that are multiples of the Fundamental
-has nothing between the lines
-the lines represent the harmonic frequencies

Front 13

If you make your voice softer or louder, you’ll see some changes in the relative strength of the ______. If you raise and lower your voice pitch, you’ll see the _____ spreading a part from one another or coming back together as the pitch comes back down again.

Back 13

harmonics, harmonics

Front 14

NOISE is when ___ frequencies are present with various ______ phase relationships to each other with approximately equal amplitude

Back 14

all, different

Front 15

NOISE (3) on a spectrum

Back 15

equal amplitude
all frequencies
random phase

Front 16

Complex periodic signals would have ______ lines.

Back 16

multiple

Front 17

simple sine wave, there would only be ___ _____ in your display

Back 17

one line

Front 18

the spectral envelope encloses an area that would have been completely filled in by those vertical lines that represent the _____ sine waves

Back 18

individual

Front 19

So the ____ ______ here is a representation of where the tops of all those lines would have been. That shows us the relative strength of the different frequency components in the noise.

Back 19

spectral envelope

Front 20

REAL VOICE SIGNALS (5)

Back 20

- the voice source is not truly periodic
- it is ‘nearly periodic’
- spectrum does not have pure lines
- spectrum has peaks
- some spread of energy around the fundamental and harmonics

Front 21

by definition if it’s not ____ in the voice, then it’s noise

Back 21

harmonic

Front 22

Harmonics to Noise Ratio

Back 22

ratio of the height of these harmonic components to the level of noise between them

Front 23

FFT SPECTRUM (4)

Back 23

-clearly shows harmonic energy (range)
- each peak is a harmonic
- less clear at showing formants
- more revealing of source

Front 24

LPC SPECTRUM

Back 24

-shows spectral envelope
- good at revealing formants
- does not show harmonics
- more revealing of filter (what its resonance frequencies are relative to one another)

Front 25

FFT stands for?

Back 25

Fast Fourier Transform
-more up and downs lines on graph (individual harmonics)

Front 26

LPC stands for?

Back 26

Linear Predictive Coding
-long horizontal line on graphs (formants)

Front 27

Formants

Back 27

resonant energy peaks in the vocal tract transfer function

Front 28

Formants don't care if you have phonated or not. (t/f)

Back 28

TRUE

This vocal tract transfer function applies whether the sound coming into it is periodic (or almost) and phonated, or whether it’s merely being whispered. The formants you see in the vocal tract will be present because of the way you have shaped the vocal tract.

Front 29

spectrogram shows us?

Back 29

individual spectral slices that are arranged side by side over time
-how the frequency components of the signal do change over time.

Front 30

A _______ shows the ingredients of a sound at a single point in time.

Back 30

spectrum

Front 31

Spectrogram is hybrid because?

Back 31

hybrid domain display since it shows both time and frequency.

shows us how the strength in each range of frequencies is changing or evolving over time.

Front 32

SPEECH SPECTROGRAM (3)

Back 32

- x-axis is time
- y-axis is frequency
- darkness indicates intensity

Front 33

SPECTROGRAM–it combines _______ and _____

Back 33

frequency, time

Front 34

The sample rate needs to be at _____ the frequency of the highest component frequency in order to accurately record and reproduce it.

Back 34

twice

Front 35

Spectrograms are made from digital recordings, and so the frequency display on the Y axis is limited to the highest frequency component you were able to save and reproduce, or the ______ _______.

Back 35

Nyquist frequency

Front 36

Examples: If you were interested in fricative energy that might go all the way up to 12-14 kHz, you’d have to make sure that the sample rate of your digital recording was at least double this amount --maybe ____-____ kHz to make sure that you include all of that interesting fricative energy. If you only care about the info from 0-4000 Hz, then you could sample at __-____ Hz. (The ____ ______ would be half your sample rate: 4-5000 Hz.) Then you would be able to display all the information that you needed.

Back 36

25-30, 8-10000, Nyquist frequency

Front 37

When you’ve made a high frequency recording, you can always _______ it later.

Back 37

downsample

Front 38

SOUND SPECTROGRAMS (3)

Back 38

• display reflects the contribution of
many structures and movements
• much detail is present for even simple utterances
• need to be selective, specific in
interpreting the display

Front 39

SPECTROGRAM PARAMETERS(4)

Back 39

• y-axis (frequency) limited to Nyquist
–actually, slightly below Nyquist
• sample at a high enough rate
–what details do you want to see?
• display can be adjusted downward
• no way to adjust upward beyond
Nyquist frequency

Front 40

A ____ band spectrographic analysis gives you good TIME, or temporal detail. It can show you individual events that are very _____, and resolve those with great _______.

Back 40

wide, brief, accuracy

Front 41

A _______ band spectrogram is a little different in that it tends to blur things together over time, but it gives you extremely good FREQUENCY resolution.

Back 41

narrow

Front 42

the laws of physics prevent us from having good ____ AND ______ detail on a 3D spectrogram display

Back 42

time, frequency

Front 43

ANALYSIS BANDWIDTH
wide vs. narrow?

Back 43

ANALYSIS BANDWIDTH
• ‘wide band’ spectrogram gives clear temporal detail
–frequency resolution is poor
• ‘narrow band’ spectrogram gives clear frequency detail
–time resolution is poor

Front 44

On a wide band spectrogram the vertical striations, or the vertical thin lines, represent each ____ _____ as it occurs in voicing.

Back 44

glottal pulse

Front 45

A___ Hz bandwidth is what we refer to as a Wide Band spectrogram.

Back 45

300

Front 46

Wide Band Spectrogram :
x and y axis

Back 46

X axis going from left to right represents TIME. It is time-aligned with the microphone signal displayed in the waveform above. Up the Y axis is the FREQUENCY.

You can see that there are faint grey lines. Each one of those represents 1 kHz (kilohertz). This one goes up to about 4 kHz on its Y axis. The lightness or darkness of the trace represents the strength of energy, or amplitude. Darker = more energy.

Front 47

This was made using a ___ Hz bandwidth –it’s a NARROW BAND display

Back 47

45

Front 48

Narrow Band Spectrogram characteristics:

Back 48

you can see horizontal, almost parallel, lines. These are the HARMONICS,which the wide-band display could not show because it couldn’t give us good frequency detail. In this narrow band display, we have extremely GOOD FREQUENCY DETAIL.We can tell the harmonics apart from one another.

Front 49

Main Idea of Wide Band Spectrogram:

Back 49

WIDE BAND: shows individual vertical glottal pulses (vertical striations) good TIME resolution

Front 50

Main Idea of Narrow Band Spectrogram:

Back 50

NARROW BAND: shows horizontal lines that represents HARMONICS of the voice, separates them out. –good FREQUENCY resolution

Front 51

A wide and narrow band spectrogram display of a normal and hoarse person. You can see by comparison that the normal voice has harmonics that go _____ up the screen. The hoarse voice’s harmonics ___ out sooner and are replace by faint noise higher up.

Back 51

higher, drop

Front 52

Function, or the way we use our ______, etc., can also contribute to our dialect or other features of our voice which are not anatomically based.

Back 52

articulators

Front 53

The larynx and vocal tract differ ________ from one person to the next. This of course influences the acoustics of the sounds the person will produce. There are structural differences that cause the _______of your own voice.

Back 53

anatomically, uniqueness

Front 54

SLPs work on altering the way the client uses his ______to improve sound output.

Back 54

structures

Front 55

_____ & ______ tend to carry the greatest amount of energy in speech

Back 55

Vowels & diphthongs

Front 56

why won’t a man’s high pitched voice
sound like a woman’s?

Back 56

–vocal fold closure patterns
–harmonic spectral slope
–noise between formants

Front 57

Women are about an octave higher pitch than men because

Back 57

–men’s vocal cords are larger

Front 58

When you produce a vowel?

On the other hand, consonants are usually produced very _____.

Back 58

The vocal tract shape can be held constant, and a vowel can be prolonged as long as you have breath –the vowel is TIME INVARIANT

briefly

Front 59

It’s the _____ __ ___ _____ _____ or the configuration of the _______that determines the particular vowel that we will be producing at that moment.

Back 59

shape of the vocal tract, articulators

Front 60

What makes one vowel ______ from another is that the frequencies of the formants are different for each one.

Back 60

different

Front 61

The output from the larynx is basically the ______ for every sound we produce.

Back 61

same

Front 62

On the ____ (what kind of display) exact locations of these formants allow us to distinguish between the different _____ that are being produced.

Back 62

LPC, vowels

Front 63

Usually you can identify the vowel from the first __ formants, even with synthetic speech.

Back 63

2

Front 64

The higher formants –F3, F4, F5 and so on –tend to make the vowels sound more ____.

Back 64

natural

Front 65

______ are peaks in the vocal tract transfer function.

Back 65

FORMANTS
In other words, as energy passes through the vocal tract, there are certain frequencies that are favored, and other frequencies that are weakened or attenuated

Front 66

The ____of a resonator will dictate what frequencies are favored (or echoed or resonated) and which are not.

Back 66

size

Ex pipes on an organ

Front 67

_____ transfer is more efficient at the formants.

Back 67

Energy

Front 68

Children have _____formant frequencies than men or women, because they have _____ vocal tract structures, including _____ resonating cavities.

Back 68

higher, smaller, smaller

Front 69

F1 and F2 plots on a graph:
the vowel quadrilateral for children is the _____ and it spreads the _____ up and to the right. This is because their formant frequencies are higher. The women’s quadrilateral is in the middle, and the men’s is the furthest down and to the left, because their formant frequencies are lower. (The men have larger resonating cavities in their vocal tracts.)

Back 69

largest, furtherst

Front 70

/i/ is a ___ ____ vowel; its first formant is rather ___.

Back 70

high front, low

because the constriction formed by the tongue in the vocal tract is forward in the mouth and high. The constriction is relatively small because the tongue nearly touches the palate.

Front 71

/ae/ (which is a ___ _____ vowel,

because the tongue is still far forward in the mouth but the constriction is quite wide) the first formant _____in frequency.

Back 71

low front, increases

Front 72

/u/ is a ___ ____vowel –so its first formant is rather ___.

Back 72

high back, low

Front 73

________THE JAW AND TONGUE INCREASES THE FREQUENCY OF THE FIRST FORMANT (F1).

Back 73

LOWERING

Front 74

AS YOU MOVE THE TONGUE _______, THE SECOND FORMANT’S (F2) FREQUENCY INCREASES.

Back 74

FORWARD

Front 75

HIGHER VOWELS = _____ F1

Back 75

LOWER

Front 76

LIP ROUNDING _______ ALL FORMANTS.

Back 76

LOWERS

lengthening the vocal tract

Front 77

LOWER VOWELS = _______ F1

Back 77

HIGHER

Front 78

FRONT VOWELS = _______ F2

Back 78

HIGHER

Front 79

BACK VOWELS = ________ F2

Back 79

LOWER

Front 80

______________ interference, where the waves will add up and give you more strength.

Back 80

Constructive

Front 81

___________ interference, where they may cancel one another out.

Back 81

Destructive