acoustics 2a

Slide1: 

Recall: Historical aside: these formulas or something like them were known for centuries by those constructing organ pipes, flutes, and the like.

Slide2: 

Circumstances under which speed of sound may vary: when deep-sea divers breathe a helium-oxygen mixture and/or air under high pressure. for entertainment purposes. [Note: it is unsafe to breathe the helium from the cans used to inflate party balloons. There is a propellant in the cans that can be unhealthy!]

Slide3: 

Circumstances where the length of the vocal tract tube varies: Age difference: children have shorter vocal tracts than adults. Sex difference: adult females have (on average) shorter vocal tracts than adult males Action of the lips and larynx: lip protrusion and larynx lowering lengthens the vocal tract (both are typically used to produce [u]) A ‘smile’ (lip corner retraction) effectively shortens the vocal tract a bit (in vocalizers with snouts, e.g., chimps, dogs, mouth corner retraction shortens the vocal tract a lot).

Slide4: 

The formula given earlier applies only to a uniform tube, i.e., one having uniform cross-dimensional area. This is true for just one vowel: [3] as in British English “bird” [b3d]. (It is also close to the lower front rounded vowels [] and [].) But most vowels do not have a uniform area function from glottis to lips. How can the resonant frequencies be calculated? There are formulas (more complex than the above equation for uniform tubes) and there are computer programs implementing these algorithms. There are also “rules of thumb” :

Slide5: 

Rules of Thumb predicting Resonant Frequencies of Non-Uniform Tubes 1. A constriction at a node (of the standing wave) raises the resonant frequency (vis-à-vis its value for a uniform tube). [Correspondingly an expansion at a node lowers the res. freq.] 2. A constriction at an antinode lowers the resonant frequency (vis-à-vis its value for a uniform tube). Both principles apply gradiently: the narrower the constriction the more the res. freq. changes; the closer the constriction is to the node or antinode, the more the res. freq. changes.

Slide6: 

In applying these heuristics, it is necessary to know where in the tube the familiar articulatory ‘landmarks’ are:

Slide7: 

What is effect of palatal constriction? What is effect of pharyngeal constriction? What is effect of labial constriction? What is effect of velar-uvular constriction? What is effect of labial + velar-uvular constriction?

Slide8: 

What are the ‘best’ vowels? If a language had only 5 vowels, would you expect them to be: A: [ i e A o u] B: [i I y Y  ] C: [y    ] Why?

Slide9: 

An important criterion for selecting vowels for a language is that they be different from each other. In all three sets the vowels are different but in some cases not as different as they might be. Set B has all the vowels crowded in the high-front region of the vowel space. It doesn’t make maximum use of the available dimensions. But what about Set C: as plotted on the vowel space, it has the same dimensions as Set A.

Slide10: 

Consider Set A: F1

Slide11: 

Effect of lip rounding

Slide12: 

The effect of no lip rounding  

Slide13: 

So: “opposite” rounding yields less distinct vowels

Slide14: 

Case Study 2: medial consonant clusters often assimilate completely to place of articulation of the second consonant Formulaically : C1C2 > C2 e.g., Late Latin oktu “eight”  Italian otto noktu “night”  notte laktu “milk”  latte Previously this was explained as a case of minimizing effort because it was assumed that if speakers found it more difficult to say two dissimilar consonants in a row than to simplify the cluster. But why is it almost always C2 that survives? One might think that if energy was being minimized the speaker would utter C1 and only then realize that energy could be saved by not uttering C2.

Slide15: 

I present a different scenario – one that looks at how such clusters are perceived by listeners.

Slide16: 

Here, when there a conflicting cues as the place of articulation of the stop ‘event’, the more auditorily salient and rich place cues at the release of the second stop dominate over those an the beginning of the first stop. The listener pays attention to the stronger cues. Again, the change arose through an innocent mistake by the listener. That is, there is no teleology in the change.

Slide17: 

The taxonomy of source-tract configurations: Sound source = Oral vowels, median approximants, the transition portion of any oral consonant. Such a configuration will produce only resonances (and formants) (also known as “poles” in mathematical terms).

Slide18: 

Nasal vowels and approximants A branched resonator gives rise to anti-resonances (and anti-formants) the spectrum. (These are also known as “zeroes”.)

Slide19: 

Nasal consonants

Slide20: 

Voiceless fricatives, stop bursts If the impedance of the constriction is high only resonances from the “front” (downstream) cavity will shape the aperiodic source; if the impedance is low enough, then the back cavity effectively makes this into a branched resonator and zeroes will be present, too.

Slide21: 

Voiced fricatives (glottal source is periodic; supraglottal source is noise pulsed at the same rate as the voicing).

Slide22: 

Laterals are often characterized as having a configuration similar to nasal consonants, the open branch being the lateral one and the ‘cul-de-sac’ being the median channel. That’s possible, but it may also be the case that the abrupt amlitude and spectraldiscontinuities of laterals is created by an abrupt change in area function: During the lateral closure Before or after the lateral closure

Slide23: 

In-class demo from Fall 2000.

Slide24: 

Acoustic Phonetics is a gas! (When speaking with different gases filling your vocal tract.) This is PowerPoint documentation of the in-class demo involving the effects on speech of speaking when one breathes gases with speeds of sound different from that under normal conditions.

Slide25: 

Background: recall the equation for determining the resonant frequencies of the vocal tract with uniform cross-dimensional area: fn is the nth resonant frequency; c is the speed of sound (34,000 cm/sec in the oxygen-nitrogen mixture of the normal atmosphere at sea level); L is the length of the vocal tract. We usually think of c as being constant and normally it is. But not always.

Slide26: 

Deepsea divers cannot breathe the normal oxygen + nitrogen mixture because nitrogen takes a long time to diffuse out of the bloodstream and back into the atmosphere. When the diver moves too rapidly towards the surface and the external pressure decreases the nitrogen forms small bubbles in the blood which can impede the flow of blood through narrow capillaries. This is the source of “the bends”, a debilitating and potentially fatal illness. To eliminate this, divers breathe a mixture of oxygen + helium. Helium does not have the same adverse properties of nitrogen. However, there is one annoying side-effect of the helium: it gives this hybrid gas mixture a much higher speed of sound and thus affects the vocal tract resonances. Sulfur hexafluoride is an industrial gas that has a speed of sound lower than normal atmospheric air. It is not normally breathed but it doesn’t harm one when it is breathed in. It does affect the voice, though!

Slide27: 

First the audio: normal voice: speech with helium + air mixture speech with sulfur hexafluoride + air mixture: [Thanks to our volunteer, Yelda!]

Slide28: 

Now the spectrograms (just the second repetition of the last two): Normal Helium Sulfur Hexa-fluoride