L105/205 – Phonetics Handout 13 Scarborough Feb. 28, 2005 reading: Johnson Ch. 3 (today); Johnson Ch. 4 (Wed) paper: By the end of this week, you should have a concrete plan for your term project.
The Auditory System
1. Remember the speech chain: (planning) → articulation → acoustics → audition → perception
(from Denes & Pinson, 1993)
• audition = hearing: registering sounds in the brain an acoustic signal in the air is converted into an electrochemical signal in the brain • perception: decoding the message an electrochemical signal in the brain is decoded into segments, words, and ultimately, meanings
2. The auditory system
(from Borden, et al., 2003)
• Outer ear ° pinna – funnels the sound (especially from in front of the head) the small projection of the pinna over the opening to the ear canal: tragus ↓ ° external auditory meatus = ear canal – protects the very delicate parts of the middle ear cerumen (wax) and cilia (hairs) line the canal and filter out dust, etc.
The ear canal also filters sound since it is a resonator tube (open at one end) - amplifies frequencies around 3500 – 4000 Hz
• Middle ear ° tympanic membrane = eardrum – a thin, stretched membrane separating the outer ear from the middle ear; vibrates with fluctuations of air pressure (sound) ↓ ° ossicular chain – a chain of 3 bones attached at the inside of the eardrum that propagate and amplify sound vibrations malleus (hammer) → incus (anvil) → stapes (stirrup)
(from Denes & Pinson, 1993)
The signal will have to be propagated next to the fluid-filled inner ear. But liquid offers a higher impedance or resistance to sound pressure than air does, so the sound pressure must be increased so that energy transmission isn’t just blocked (reflected). The ossicular chain converts small vibrations on a large surface (tympanic membrane) → large vibrations on a small surface (oval window). Sound pressure is increased by approximately 30 dB. ↓ ° oval window – a membrane leading to the inner ear
• Inner ear ° cochlea – a fluid-filled, coil-shaped duct in the temporal bone of the skull; contains the basilar membrane and the organ of corti Rocking of the stapes in the oval window is translated into pressure variations in the cochlear fluids. ↓ basilar membrane – coiled membrane stimulated by movements in the cochlear fluid
“tonotopic” organization ° each piece of the membrane responds to different freqs • where – frequency • how big – loudness → gives a spectrum of the sound
(from Borden, et al., 2003)
The membrane is narrow and stiff at the base and wider and less stiff at the apex (opposite of what one might expect). ↓ organ of corti – actual sense organ of hearing (auditory receptor); holds rows of hair cells against the basilar membrane and releases an electrochemical signal to the auditory nerve
• Auditory nerve (= 8th cranial nerve) – a bundle of nerve fibers coming from the hair cells ↓ exits the temporal bone via the internal auditory meatus; passes through the brainstem ↓ auditory cortex – area in the temporal lobe (Heschl’s gyrus) The frequency and intensity information are represented directly (topographically) in the temporal lobe.
3. Sound begins as acoustic energy. → mechanical energy as it hits the tympanic membrane (and is transmitted through the ossicular chain) → hydrodynamic energy in the cochlea → electrochemical energy as hair cells are activated via the basilar membrane
Loudness 4. The amplitude of a sound wave = the displacement of air pressure from atmospheric pressure. - usually measured as root-mean-squared (RMS) amplitude i.e., the square root of the mean of the squared amplitude Why? A sound wave involves both negative and positive displacements from atmospheric pressure. Amplitude is the magnitude of displacement, either positive or negative. Squaring makes all the displacements positive so we can take the mean, and then we undo the squaring by taking the square root.
5. The auditory sensation of loudness - The perceived loudness of a sound depends on its amplitude. But subjective auditory impressions of loudness differences do not match sound pressure differences. e.g., For soft sounds, small changes in pressure yield large changes in perceived loudness; for loud sounds, large changes in pressure yield small changes in loudness. e.g., Loudness is sensitive to frequency: low and high frequencies are perceived as quieter than mid frequencies.
- sones: units of perceived loudness based on subjective judgments
(from Johnson, 2003)
- decibel (dB): unit of relative loudness that provides an approximation of the nonlinearity of human loudness sensation; measured in terms of intensity • Intensity is proportional to the square of the amplitude. • We can measure the intensity of one signal relative to another: the intensity of a sound x relative to a reference sound r : x2/r2 2 2 • A bel is the base 10 log of this ratio : log10(x /r ) 2 2 A decibel (dB) is one tenth of a bel : 10 log10(x /r ) = 20 log10(x/r) • Reference level: - 20 µPa : lowest audible pressure fluctuation of a 1000 Hz tone (dB SPL) - lowest audible pressure fluctuation at that frequency (dB SL) - could be specified as anything else (Note that the log scaling of the decibel scale approximates human perception of loudness by representing enhanced sensitivity to differences in sound pressure at low pressure levels. However, the decibel scale actually exaggerates this non-linearity.)
Frequency response 6. The response of the auditory system to frequency is also non-linear. e.g., For low frequency sounds, changes in frequency are perceptually greater than acoustically equivalent changes for higher frequency sounds. i.e., The auditory system is more sensitive to frequency changes at the low end of the audible frequency range than at the high end.
- This effect is due to the physical structure of the basilar membrane. A relatively large proportion of the membrane responds to low frequency sounds, so frequency resolution is much better in this range.
(from Johnson, 2003)
- Bark: units of auditory frequency
(from Johnson, 2003)
7. Auditory representations Since acoustic scales of loudness and frequency do not match auditory scales, acoustic analyses of speech may not accurately represent what a listener experiences. - To avoid this mismatch, we can implement a functional model of the auditory system in our analysis.
- Bark and sone scaling are examples of one-dimensional models. - Spectra can also be calculated according to an auditory model.
acoustic spectrum auditory spectrum (from Johnson, 2003)
• Auditory models allow us to look at the speech signal from the listener’s point of view.