bbn--ang--141 foundations of phonology phonetics 3

BBN–ANG–141 Foundations of phonology

Phonetics 3: Acoustic phonetics

Zoltán Kiss

Dept. of English Linguistics, ELTE

25 March 2009

z. kiss (elte/delg) intro phono 3/acoustics 25 March 2009 1 / 60

outline

sound waves

frequency

amplitude

synthesis

harmonic analysis

types of sounds

the source–filter model

formants

spectrograms

sample exam questions

recommended reading


sound waves

acoustics

definition and etymology

I acoustics is a branch of physics and is the study of sound (which ischaracterized as mechanical waves in gases, liquids, and solids)

I acoustic is derived from the Greek word Ćkoustoc ‘able to be heard’

I it is concerned with the production, control, transmission, reception,and e=ects of sound

I it aims at describing and quantifying the properties of sounds with thehelp of various wave-related models

acoustic phonetics

deals with the acoustic properties and quantification of speech sounds


sound waves

what is sound?

whenever there is a sound, there is:

sound transmission

I sound source

I transmission through a medium (e.g., air, water)

I potential receiver/interpreter

the definition of sound

Sound is a potentially audible disturbance of a medium produced by avibrating source.


sound waves

two problems:

I sound is invisible

I most sounds are fairly complex

the task:

I make sound visible for analysis

I deal with the simplest sounds first


sound waves

the simplest sounds: pure tones

the tuning fork emits pure tone


sound waves

the sound of the tuning fork


sound waves

simple harmonic motion (SHM)

the SHM of the tuning fork

I the tines of the tuning fork vibrate in simpleharmonic motion

I the tines move back and forth a fixednumber of times per second (no matterhow hard the fork is struck)

I periodic motion: the pattern repeats itselfuntil it damps out


sound waves

simple harmonic motion (SHM)

the SHM of the tuning fork

I a complete movement betweenstarting/rest position > maximumdisplacement > back over starting position> maximum displacement > back to startingposition = a cycle (c)

I frequency (F): the number of completedcycles per second (s) (Hertz (Hz) or cps)

I the tines complete 440 cycles per second,frequency of the tuning fork = 440 Hz


sound waves

simple harmonic motion (SHM): the swing &

the pendulum of the grandfather clock


sound waves

the propagation of sound: pressure wave movement


sound waves

sound propagation (of the pure tone): summary

I SHM of sound source

I SHM of air particle set in motion by source

I air particle moves in sympathy with the SHM of source

I individual particle has limited motion

I areas of air compression and rarefaction /re@rI"fækSn/ are created

I compression and rarefaction areas move in time away from source,transmitting the SHM of source (pressure wave movement)

I listener senses same SHM as that of the source: sound has beenpropagated


sound waves

?how can we graphically represent the SHM of the air particles?


sound waves

representations of sound propagation: waveform


sound waves

waveform

A waveform is a display of how amplitude varies over time.


sound waves

from SHM to waveform: movement-based graph


sound waves

from pressure wave to waveform: pressure-based graph

– variations in air pressure with respect to an equilibrium /i:kwI"lIbrI@m/


sound waves

pure tone (sinuosoid) waveform

– variations in air pressure with respect to an equilibrium /i:kwI"lIbrI@m/


sound waves

waveform anatomy

simple harmonic motion

I simple: has only one frequency; harmonic: can be graphed as a sinus wave

I periodic: the cycles repeat themselves until they damp out

I its waveform is characterizable by two independent parameters:

1. frequency (above: F = 100 Hz) or period (T; the time for a cycle tocomplete in seconds; above T = 0.01 s)

2. (peak) amplitude (the distance from the zero crossing)


frequency

change in frequency ⇒ subjective sensation of pitch


frequency

some important facts about frequency

I when a sound is twice the frequency of another sound, it is an octavehigher

I frequency range of human hearing: 16–20 Hz–20,000 Hz

I speech sound analysis usually involves the range between100 Hz–10,000 Hz


frequency

change in amplitude ⇒ subjective sensation

of loudness/intensity


amplitude

the decibel: a measure of relative intensity

why the decibel scale?

I air pressure amplitude is measured in pascals (Pa)

I the pascal scale is a linear scale: each increment is equal to the next

I the sensation of sound loudness/intensity is related to amplitude;however,

I it is not linear but logarithmic, that is,

I it is constructed with increments with increasingly large numericaldi=erences

I the decibel (dB) scale (or “sound pressure level (SPL) scale”) is alogarithmic scale of the amplitude of air pressure variations

I the dB scale has intervals that are roughly equal to perceived loudness


amplitude

0 dB = 20 µPa(the threshold of hearing; the buzz of a mosquito around 3 meters away)


amplitude

80 dB (≈ 100000 µPa)(average street tra;c)


amplitude

140 dB (= 100,000,000 µPa)(threshold of pain; jet engine at 25m distance)


synthesis

?what about complex sounds?!

(speech sounds are nothing like pure tones!)


synthesis

how can we characterize complex waves?

I key idea: if we can reduce a complex periodic waveform into acombination of sine waves

I then we can describe it using information about the frequency andamplitude of each component sine wave


synthesis

Jean Baptiste Joseph Fourier(1768–1830)


synthesis

to build a complex wave is like a recipe, e.g., take

I 1 100 Hz/30 dB sinus wave, then add

I 1 200 Hz/10 dB sinus wave, and also add

I 1 300 Hz/20 dB sinus wave

This addition of two or more di=erent sine waves to create a complexperiodic wave is called synthesis.


synthesis

waveform of a complex tone derives from 2 or more pure

tones of di=erent frequency and/or amplitude


synthesis

three important consequences of synthesis

I the amplitudes of the complex wave depends on the addition of theamplitudes of the component waves

I the sine wave with the smallest frequency will define the main/basicrepetition frequency of the complex wave: fundamental frequency f0

I the other sine wave frequencies present in the complex wave are calledharmonics (H) (or: overtones);

harmonics and f0

harmonics are integer (whole number) multiples of the f0(this is because each sine wave component must complete a whole number of cycles within

one period of the complex)


harmonic analysis

Our example complex wave has this harmonic series (also called Fourierseries):

Harmonic Frequency Amplitude

H1 (= f0) 100 (100 × 1) Hz 30 dBH2 200 (100 × 2) Hz 10 dBH3 300 (100 × 3) Hz 20 dB

harmonic analysis

the reverse of synthesis, finding (characterizing) the component sine waveharmonics of the complex wave


harmonic analysis

Fourier’s theorem (1822)

I All complex periodic waveforms can be analysed into a sum ofsinusoidal component waveforms (harmonics).

I The mathematical algorithm of this process of harmonic analysis iscalled Fourier analysis or Fourier transformation.


harmonic analysis

?how can we graphically represent harmonic analysis?


harmonic analysis

spectrum graphs

I the (power/amplitude/line/sound) spectrum (plural: spectra): is aplot of the results of harmonic analysis

I frequency of harmonic: horizontal axis

I amplitude of harmonic: vertical axis

I time and phase: not shown (Fourier analysis is taken at a particularinstant of time)


harmonic analysis

harmonic series of D

harmonic freq. ampl.first (=f0) 100 Hz (100× 1) 30 dBsecond 200 Hz (100× 2) 10 dBthird 300 Hz (100× 3) 20 dB


harmonic analysis

loudness, pitch, quality: a summary

I sound loudness depends on amplitude

I sound pitch depends on f0

I sound quality/timbre depends on the spectrum (harmonic series)

I all these components are independent of each other


types of sounds

types of sounds based on their acoustics

I periodic/repetitive (with a clear sense of pitch)I simple: with only one frequency

(sine waves/pure tones, the tuning fork)I complex: 2 or more component frequencies that are harmonically

related: f0 + higher harmonics(vowels, sonorants: e.g., [j], [w], [l], [r], nasals)

I aperiodic/nonrepetitive (less pitch, sense of noise)I complex: 2 or more component frequencies that are not harmonically

related (the number of harmonics is “infinite”)I transient/impulsive aperiodic sounds (noise bursts, taps, clicks, e.g., the

release of plosives (e.g., [p], [t], [k]))I continuous aperiodic sounds (white noise, fricatives: e.g., [s], [S], [f])


types of sounds

the prototypical aperiodic sound: white noise


types of sounds

last week. . .

I you heard about the articulation of vowels

today. . .

I what are the most important acoustic properties that characterizedi=erent vowels?

I how are articulation and these acoustic properties related?



the basic acoustics of vowels

resonators and filters

I each object vibrates when excited (frequency response), and so it actsas a resonator (system)

I acoustic resonators: objects that contain body of air that is set intovibration by a source

I each resonator vibrates at maximum amplitude at a preferred specificfrequency range, the natural frequency (fn)

I the fn of a resonator largely depends on the shape and size of theobject (cf. violin vs. cello)

I the resonator acts like a kind of filter as it lets through frequencies witha greater amplitude near its fn



what do filters do?

input source signal

⇓

resonator filter — modifies input frequency amplitude

⇓

(modified) output signal



frequency response curve/graph of a resonator



the basic acoustics of periodic sounds

the source–filter model of periodic speech sounds (e.g., vowels)

I the production of periodic speech sounds can be modelled with theinput source + resonator filter model

I source: the periodic vibrations generated by the vocal folds in thelarynx (f0 or fx or Lx ; resembles a buzz)

I filter: the vocal tract cavities (pharynx, oral cavity, and nasal cavity)which modify the shape of the input source

I output: the actual speech sound uttered (“final acoustic product”)



typical characteristics of the source in the larynx

I its intensity is set by lung pressureI its f0 (also called fx – larynx for speech sounds) is set by the tension and

length of the foldsI typical f0 for men: 100–200 Hz; for women/children: 150–350 HzI its spectrum shows a gradually decreasing slope towards the high

frequencies



typical characteristics of the vocal tract filter cavities

I each of the vocal tract cavities above the larynx acts like a complexresonator/filter

I thus each can be considered as if made up from a number of simpleresonators strung together

I these complex resonators will have several response curve peaks



⇓

the sound in bird [3:]


formants

so: what are formants?

formants are frequency peaks in the output signal as a result of the shapingof the input source by the vocal tract resonators as complex filters

some basic information on formants

I formant frequencies depend on the position of articulators, like thetongue

I the constant movement of the articulators during speech createvariable vocal tract resonators, hence variable formants

I frequency of the first 2 or 3 formants (F1, F2, F3) are the mostimportant for setting phonetic quality and phonological contrast

I F1 ∼ pharynx/throat; F2 ∼ oral cavity

I e.g., the vowel in bird [3:] can be characterized by having formants at:F1 = 500 Hz, F2 = 1500 Hz, and F3 = 2500 Hz


formants

vocal tract frequency response peaks for [i], [A], [u]


formants

correspondence between tongue position and formants

I front vowels have high F2

I low vowels have high F1

I rounding lowers F1 & F2


spectrograms

how can we represent spectra in “real time”?

I we can use spectrograms

I they are series of spectra in quasi-3D

I they encode information on:I time (in seconds) – axis xI frequency (in Hz) – axis yI amplitude (in dB) – greyscale


spectrograms

spectrograms of bead & barred


spectrograms

spectrograms of sigh & shy


spectrograms

Peter Ladefoged(1925–2006)


spectrograms

sound classes and how they typically look on spectrograms

I vowels: steady state, clear, intense formant structure at typicalfrequencies

I fricatives: no clear formant structure, wide range of frequencies, butfrequency groupings at specific frequency ranges (darker areas inspectrogram)

I plosives/stops: no frequencies during closure phase (white area inspectrogram), can have short noisy phase after release

I sonorant consonants: similar spectrograms to vowels but the formantstructures are less clear (esp. for nasals)

I place of articulation: is signalled by formant transitions (falling, risingformants), formant shapes (e.g., bet ⇐⇒ get)

I voicing: is signalled by intensity (darker areas) at very low frequencies,below F1 (this intensity actually shows the presence of thefundamental frequency (f0))


spectrograms

that’s all! but before you go. . .




which of these is pitch related to?

1. fundamental frequency

2. amplitude

3. formants

4. intensity

which vowel has the highest F2 value?

1. [u]

2. [i]

3. [a]

4. [O]


recommended reading

Some recommended reading (not compulsory for the exam!)

I Ladefoged, Peter (2005): Vowels and consonants. Second edition.Malden MA/Oxford, Blackwell.

I Ladefoged, Peter (1996): Elements of acoustic phonetics. Secondedition. Chicago/London: The University of the Chicago Press.

I Borden, Gloria J., Katherine S. Harris and Lawrence J. Raphael (2003):Speech science primer. Physiology, Acoustics, and Perception of Speech.(Fourth edition). Baltimore: Lippincott Williams & Wilkins.


bbn--ang--141 foundations of phonology phonetics 3

Documents