22 Loudness ME explained

Loudness

perception based (different than measured dB)

the subjective magnitude of a sound

the attribute of auditory sensation in terms of which sounds can be ordered from soft to loud

Subjective magnitude

In detection and discrimination experiments

• subjects judgments are either right or wrong

• described in terms of performance (5 of correct responses for a given condition)

• loudness tasks

• no correct or incorrect response

• a subjective task

Quantifying loudness

matching tasks

loudness of sound in 1 ear is compared to loudness of sound in other ear

quantifying loudness: one ear

standard tone; fixed level

quantifying loudness: other ear

comparison tone; subject has control over its level & adjusts the level until the subjective magnitude is equal in both ears

loudness matching: normal hearing

expected outcome w/ same frequency pure tones in each ear & normal hearing (will get slope of 1)

x axis = level of tone in left ear comparison tone

DB SPL

y-axis = level of tone in right ear (standard tone dB SPL)

loudness metering: conductive loss

slope of tone

simply displaces the function by the amount of loss

x- axis: level of tone in left ear (conductive loss) comparison tone (dB SPL) (ex. middle ear filled w/ fluid)

y-axis - level of tone in right ear (normal standard tone) (dB SPL)

Loudness matching: cochlear loss

abnormally rapid growth of loudness

metered to other hearing loss

outer hair cells are damaged

sound is elevated

x - axis = level of tone in left ear cochlea loss, comparison tone (db spl)

y - axis: level of tone in right ear (normal)

standard tone (db spl)

loudness requirement

an abnormally rapid growth of loudness seen in persons w/ cochlear

loudness recruitment can be stimulated by shifting a normally hearing persons threshold using a noise masker

loudness across frequency - phons

the level of an equally loud 1.0 khz tone

subjects are asked to match the loudness of a tone at some frequency to that of a 1.0 KHZ tone

phon curves

reduced dynamic range for low frequencies

high levels = flakes out follows audibility curve

comparison tone frequency

phons

follows audibility for low level tones

the function fluctuates at high levels

scales

matching tells us about the relative growth of loudness

psychologists interested in a scale of loudness

scale of loudness would allow us to predict loudness for sounds based on their physical characteristics

• spectrum, dB SPL, or duration

scales continued

a method of assuming numerals to objects or events

types of scales

normal scale

ordinal scale

interval scale

ratio scale

nominal scale

represents the most unrestricted assignment of numerals

numerals are used only as labels & words would serve equally well

eg. numbering of football players, code #’s for posting of grades

meaningful operation determination of equality

Ordinal scale

arises from the operation of rank ordering (eg pleasants of odors, quality of leather, wool,, lumber)

ordinal scale (meaningful operation)

determination of greater of less than

interval scale

quantitative scale w/ an arbitrary zero or reference point.

difference between scale values are meaningful. However, ratios of scale values not meaningful due to arbitrary reference point

• eg. I.Q Ferinheight, Celsius, temperature scales

Interval scale - meaningful scale

detection of equality intervals of intervals or differences

Ratio scales

these are the scales most commonly encountered in physics

the scales have absolute zero

• eg length, weight, kelvin (temperature scales)

• meaningful operations: equality, rate order differences, ratios

early attempts at scaling perceptions - fechner

believed that you could derive a scale by summing jnds

this indirect method didn’t work

JND for intensity is 1 db

If you added all of the up it would be equal, which doesn’t work

SS stevens: magnitude estimation

want to know the loudness of a sound? Ask a listener (direct approach)

the subject is asked to assign a # proportion to the apparent stimulus magnitude

goal is to produce a ratio scale

magnitude estimations

used in many experiments where the end point are known prior to the experiments

studied extensively in loudness: stimulus level easily controlled & varied over a wind range

Instructions you will be presented w/ a series of a stimuli in the irregular order

your task is to tell how intense they seem to be by assigning numbers to them. Call the first stimulus only # that seems appropriate to you.

The assign successive #’s such a way that they reflect your subjective impression for example, if the first stimulus is 20 times as intense as the first stimulus, assign a # 20x as large as first. If it seems large at first. If it seems 1/5th as large etc. you may use a whole #’s, fraction =, decimals. Try to make each # match intensity as you perceive it

magnitude estimation

10 db doubles

every 10dB = doubling of loudness

magnitude estimation demo

5 tones in quiet

6 tones in noise

magnitude estimation demo

quiet: 30 (3), 45(2), 60 (5), 75(4), 90 (1)

noise: 40 (5), 50(3), 60(2), 70(4), 80 (1), 90 (6)

class ME: response from a previous class

y = 2.1703 e^0.0418x

physiological correlate to recruitment

effect of noise o cochlear loss to estimate the active process

reduced gain by 30 to 50 dB & causes the system to become linear (ear) (30 to 50 dB hearing loss)

Schlauch, Diglovarnt (Ries) (1998)

compared basilar membrane 210 data to loudness data to see if there is a linearized basilar membrane response that can predict loudness recruitment e

Physiological correlate to recruitment

effect of noise or cochlear loss is to estimate the active process

reduce gain by 30 to 50 dB & causes the system to become linear (30 - 50 dB hearing loss)

Schlauch, DiGiovani & Rise (1998)

Compared basilar membrane 210 to loudness data to see if there is a linearized basilar membrane response that can predict loudness recruitment

nonlinear BM response: Ruggero

x - dB SPL

y - velocity (MMS)

Mathematical analysis of BM 210 functions

results showed that a linear BM response can predict loudness recruitment

Our results suggest that linearity occurs when the loss-reaches roughly 50 dB

Magnitude estimation: a ratio scale?

Replicate many times, but results change depending on the instructions

*people are inaccurate in the way they use #’s)

because of this problem, magnitude estimation produces an ordinal scale

w/ different reference values (starting points for the same stimulus, the results are not parallel as predicted by ratio scale (10dB)

Absolute magnitude estimation

subjects assume a natural context for their judgments

no exceptions are given

no way to test the accuracy of the scale produced

Types of magnitude estimation - Traditional ME

examples of ratio given in the instructions

in another validation, subjects are told also that the # corresponding to a particular stimulus (from example, assign so the first tone you hear)

absolute ME

• no reference

• no examples

• no way to assess whether participants are using #’s accurately

Magnitude estimations despite it’s problems

ME is the best method for deriving a scale of loudness

• for persons w/ hearing w/in normal limits loudness doubles for 10 dB increases in level

• Recognize that ME does not represents a direct mapping to sensory receptors

• a persons ability to accurately use of #’s plays a big role

Category rating

subject assigns one of a fixed #’s of categories to the loudness

categories can be numerical or verbal labels

goals: interval scale

example: rate on a scale of 1 to w/ “1” being “very soft” & “7” being “very loud”

Problems

subjects inaccurate in its assigning of #’s

• end points need to be defined

• - subjects reluctant to use the endpoints

• advantage: easier for naive subjects to use

• delayed auditory feedback

Category rating: effect of stuttering therapy

Martin, her old son & triden (1984)

naturalness of speech (highly natural to a highly unnatural)

scaled the naturalness of speech samples of stutters before and after the therapy w/ delayed auditory feedback

control group of people w/ no speech problems

conclusions

delayed auditory feedback reduced dysfluencies

speech remained unnatural (compared w/ normal group) even w/ low #’s of dysfluencies

Could have drown the same conclusions w/a magnitude estimations task (but more difficult for naive subjects to employ)

annoyance

quality of unwantedness of a sound

method of measurements is scaling

large individual differences, but trends across subjects

physical characteristics of sound

louder the sound annoying it is

sounds that startle tend to be more annoying

factors important to annoyance

situational

cognitive

social

Inference

more annoyed when listening for something in the presence of noise

listener location

will tolerate more noise:

in airplanes than in buses

outdoors than indoors

corresponds to the amount of attenuation of the walls

significance of the sound

aircraft

more significant if a recent crash

small town’s economic well being

Time of day

more annoying at night

lower level of background noise

how is loudness affected by background noise?

does sound interfere w/ sleep

socio-economic status

higher socio-economic status more likely to complain

• the difference is annoyance or difference in readiness to respond

national differences

national difference in the tolerance for unwanted sounds

study of swedes, British & Italians

Swedes less tolerant than British

British less tolerant than the Italians