SLHS 302 Final Review

perception

the process of interpreting input from sensory receptors in order to determine what is out there in the world

hearing

objects cause vibrations in air which travel and are captured by the ears

perception is hard

• Feels effortless which disguises difficulty of process

• Ears help capture and process sensations but they don’t interpret it

Perceptual problems are usually ill-posed

• Aims to make sense of sensory input

• Ill-posed: not enough information to uniquely determine the answer to a problem

• Our perceptual systems usually arrive at a single unambiguous interpretation of a stimulus

• Illusions give us glimpses when perception is challenged

• Interpretations can differ across individuals

• Brain gets a single signal that is a combination of individual sounds

• Somehow we can distinguish and hear specific sounds

Perceptual systems must be invariant

• Invariance: a function, quantity, or property which remains unchanged when a specified transformation is applied

• Human speech is highly variable

• Human speech recognition is remarkably invariant

Perception is inference

• Inference: the brain’s best guess about the current state of the world based on all other information it has learned about the world

• We are not aware of the assumptions our brains make

• We use the structure of the world and use that to help us solve complicated perceptual problems!

• Illusions help demonstrate that perception involves inference by showing us instances where our inferences are wrong

• Show us we are making guess about the world

• Illusory continuity appears to be an unconscious inference about what is most likely to happen during the noise

invariance

a function, quantity, or property which remains unchanged when a specified transformation is applied

inference

the brain’s best guess about the current state of the world based on all other information it has learned about the world

psychophysics

a branch of psychological sciences devoted to studying the relation between stimulus and sensation/perception

psychoacoustics

determines the relationship between listeners’ behavioral responses to sound and the physical properties of sound

psychoacoustics

Help us characterize our perceptual experiences of sound and understand the constraints of the physical auditory system itself

• Make inferences about how the ear processes sound

• Attempt to relate these inferences to the anatomy and physiology of the auditory system

psychoacoustical methods

tasks/procedures that measure detection or discrimination can be used to describe the limits of the auditory system

psychoacoustical methods

absolute thresholds, just noticeable difference, psychometric functions

absolute threshold

smallest value of a stimulus parameter that a listener can detect

Just noticeable difference (JND)

the smallest change in a stimulus parameter that can be perceived by the listener

Just noticeable difference (JND)

difference limen or difference threshold

psychometric function

used to determine “threshold”

threshold

the stimulus parameter producing a criterion proportion of “YES” responses

psychometric functions procedure

systematically change a stimulus parameter
ask a listener to respond “YES” if the stimulus is detected

Audibility thresholds

a plot of just barely audible tones of varying frequencies

MAP

minimum audible pressure

MAF

minimum audible field

range of human hearing

20 to 20000 Hz

best sensitivity range for humans

~500-8000 Hz

selectivity diminishes

below ~500 Hz & above ~8-10 kHz

why does sensitivity diminish?

Due to middle-ear transfer function

• Stiffness limits transmission at low frequencies

• Mass limits transmission at high frequencies

perceptual dynamic range

~120 dB

single neurons dynamic range

limited to ~30-50 dB

softest sound a human can detect

-10 dB

human level for discomfort/pain

~120 dB

clinical audiograms

routine measures of audibility thresholds

clinical audiograms measurement

dB HL

dB HL

[listener’s threshold in dB SPL] - [ANSI standard absolute threshold in dB SPL]

normal limits in clincial audiograms

0-25 dB HL

audiograms

plotted “upside down” so “worse” thresholds are lower on graph

Average audiograms across the life span show

that hearing sensitivity gets worse with age

clinical audiograms

Measure hearing sensitivity thresholds; measure hearing sensitivity thresholds

hearing thresholds

do not fully capture or accurately represent all aspects of auditory perception; hidden hearing loss

A particular hearing threshold

may correspond to multiple types or amounts of damage in the cochlea or brain–a specific “site-of-lesion” cannot be identified from the audiogram alone

Hierarchy of Auditory Skills

Detection → Discrimination → Identification → Comprehension

absolute thresholds

allow us to measure hearing sensitivity

absolute thresholds

vary by tone frequency and sound duration

most sounds

heard at suprathreshold sound levels

JND measures

measures hearing abilites for audible sounds

temporal integration

the process by which a sound at a constant level is perceived as being louder when it is of longer duration

temporal integration

Suggests energy is summed over time to detect sounds more effectivelyt

temporal integration

Must consider Energy (E) and Power (P) when measuring detection of sounds with different Durations (T)

temporal integration equation

P = E/T or E= PxT

increase of 3 dB for temporal integration

P = E/T or E = PxT

as sound gets longer

it becomes easier to hear up to 300 msec

shorter sounds

need to be more intense to reach energy threshold for detection

increasing duration above 300 msec

does not make it easier to hear

Weber’s Law

the Weber fraction is often the same (constant) for all values of a physical parameter to be discriminated; fails at extreme frequency values

Weber’s Fraction

change in W / w = c

Weber’s Fraction

The JND is proportional to the smaller weight value; the constant proportion of the initial stimulus magnitude

Best intensity JNDS

~1 dB for sound levels above ~20 dB SPL

SAM (sinusoidally amplitude modulated)

used to study sensitivity to modulations directly

SAM

hard to measure due to side bands

3 frequency components of SAMs

carrier component at Fc and 2 side bands

Temporal Modulation Transfer Functions (TMTF)

Measure our sensitivity to modulation depth for different modulation frequencies

TMTF

Limit how fast modulation can be and still be detected; low-pass shaped

TMTF

faster modulations are harder to hear

maskers

sounds that interfere with the signal

signal

the sound we are trying to hear

masking

occur when the presence of a second sound interferes with the perception of the target sound

Types of masking

simultaneous masking, foward masking, backward masking

pure tones close togehter in frequency

mask each other more than tones widely separated in frequency

A tone masks tones of

higher frequencies more effectively

The greater the intensity of the masking tone

the broader the range of frequencies it can mask

Masking of tones by broadband noise

shows an approximately linear relationship to tone threshold and noise level

Most findings are explained by

BM response properties to tonal signals

BM response properties

asymmetric traveling wave; bandpass filter

Tone-on-Tone Masking

Used to measure psychophysical tuning curves (PTCs)

signal for tone-on-tone masking

low-level tone plated at a specific frequency

masker for tone-on-tone frequency

another tone is played at single frequency

PTC

a frequency map of the masker sound levels need to mask a fixed signal

PTC

similar to neural tuning curves; tells us masking occurs for both neural response and perception

Perceptual masking

constrained significantly by physiological tuning that begins in the cochlea

Masking pattern

a frequency map of the signal sound levels that are just detectable in the presence of a fixed masker

masker for masking pattern

tone with a fixed frequency and level

signal for masking pattern

tones of different frequencies

Masking is much stronger for signals at frequencies

above the masker than below

stronger masking for frequencies above the masker

Explained by asymmetry of traveling wave on the BM; More masking when masker is below signal; Less masking when masker is above signal

Signal frequencies very close to the masker

easier to hear than frequencies slightly further away

Signal frequencies very close to the masker are actually easier to hear than frequencies slightly further away

Explained by acoustics of signals involved

Beat perception is observed for harmonics of masker frequency

Explained by nonlinearity of auditory system

noise masking

energetic masking

noise masker

Must consider tone signal and spectrum level of

Signal-to-noise ratio (SNR)

key factor in tone detection in noise

The Power Spectrum Model theory of masking

listeners pick the most advantageous filter to listen through for a given signal

Broadband noise

fills up the perceptual auditory filter so the filter BW matters

Narrower noise

has all of its power within the auditory filter so noise BW matters

Critical band of frequencies

frequencies within the auditory filter are “critical” for masking

White Noise

broadband noise; a continuous and flat amplitude spectrum over its bandwidth

Perceptual Auditory Filters

the bandwidth can be estimated by measuring tone detection in noise with various psychoacoustic paradigms

perceptual auditory filters

critical raio, critical band, equivalent rectangular bandwidth

Critical ratio (CR)

measure thresholds for tones in broadband noise

critical ratio

• Detection occurs at a fixed signal-to-noise ratio

• Common for equal signal (Ps) and noise power (Pn) or SNR = 0 dB

• Estimated as 10log(BWaf) = Ps (dB) - N0 (db/Hz) or the critical band

Critical band (CB)

measure thresholds for tones in narrowband noise

critical band

• Noise BW is smaller than auditory filter BW, masking will increase as noise BW gets bigger until they are equal

• If noise BW is greater than auditory filter BW, masking will be constant

• Estimated experimentally by finding the noise BW at which further increases do not increase masking