Loudness: Current Knowledge and Questions - ITU · International Telecommunication Lannion, France, 10-12 September 2008 Union Loudness: Current Knowledge and Questions Sabine Meunier,

InternationalTelecommunicationUnionLannion, France, 10-12 September 2008

Loudness: Current Knowledge and Questions

Sabine Meunier,Laboratoire de Mécanique et d’Acoustique – CNRS - France

ITU-T Workshop on"From Speech to Audio: bandwidth extension,

binaural perception"Lannion, France, 10-12 September 2008

InternationalTelecommunicationUnion

Lannion, France, 10-12 September 20082

LoudnessSupraliminary sensation: how to measure it ?

Weber-Fechner law, 19th century

Weber: ΔΦ/Φ=constant Φ:stimulation

Fechner:ΔΨ=k ΔΦ/Φ Ψ: sensation

Ψ = A LogΦ + B

Stevens law: direct measurement, magnitude estimation

Ψ = a Φb



Methods to measure Loudness

Magnitude estimation: loudness in sones1 sone = loudness of a 1-kHz tone at 40 dB SPL

Adjustement (loudness matches): loudness level in phonsa loudness level of a sound of X phons means that the sound is asloud as a 1-kHz tone at X dB SPL

Adaptive (2down–1up, 1down–2up): loudness level in phons

Multitracking: loudness level in phons

Categorical loudness scaling



Test sound (T): the sound that we want to know the loudness

Sound of comparison (C): usually 1-kHz tone, variable level

C after T (1st test) and T after C (2nd test)

Listener’s task: Adjust the C-level to have the same loudness as T

Random order, different for each listener

Start level was randomly X dB above or under T loudness level

Loudness level in phons : mean of C-level obtained in the 2 tests for each sound

4

Methods to measure LoudnessAdjustment






Adaptive (adown–bup, bdown–aup): loudness level in phons





2down-1up (1st test) :

++

++ ++

- -

-5 dB

-5 dB+5 dB

-2 dB

1st reversal

2nd reversal

2up-1down (2nd test) :

+ +

-- --

--

+5 dB-5 dB

+5 dB

+2 dB

1st reversal

2nd reversal

+ JL

- JS

Loudness level in phons

Track finishes after X reversals

Presentation order : T & C or C & T, randomlyListener’s task : Which of these 2 sounds is louder

Start level above the T loudness level

Start level under the T loudness level

Mean of the last Y reversals

Methods to measure Loudnessadaptive methods











Principle similar to the adaptive method

4 or 5 simultaneous sequences

Random choice of the sequence, different for each listener

Loudness level in phons:

mean of +JL and –JS for each sound

Methods to measure LoudnessMultitracking











Methods to measure LoudnessCategorical loudness scaling

50

45

40

35

30

25

20

15

10

5

0

very soft

inaudible

soft

medium

loud

very loud

too loudLoudness measured in Categorical Unit (CU)



Loudness as a function of

0 20 40 60 80 100 120

11

Loudnes

s in

sones

20010050

2010

5.0

2.01.00.5

0.20.1.05

.02

Sound pressure level (dB)

8000 Hz

100 Hz

4000 Hz

1000 Hz

250 Hz

500

Hz

From Scharf (1978) in Handbook of perception, Carterette and Friedman

N=k(P-P0)a

at 1 kHz: a=0.6

N: loudnessP: pressureP0: constant

SPL




Loudness as a function of SPL

Partial loudness50

20

10

5.0

2.0

1.0

0.5

0.2

0.1

Loudnes

s in

sones

SPL of masked tone (dB)20 30 40 50 60 70 80 90 100 110

Quiet 50 60 70 80 90 100 dB SPL of Noise




Equal loudness contoursStandard: ISO 226, 2003

frequency (Hz)

sound p

ress

ure

lev

el (

dB)

frequency




frequencyEqual loudness contours

40

50

60

70

80

90

100 1000 10000

ISO226 2003ISO226 1987

Soun

d Pr

essu

re L

evel

(dB)

Frequency (Hz)

75 phons

60 phons

50 phons

From Boullet (2005)

PhD Thesis


Lannion, France, 10-12 September 200815Bandwidth (Hz)

Loudnes

s le

vel in

phons

overall sound presure level


Spectral loudness summation


bandwidth

critical band



Temporal loudness summation

16


56

58

60

62

64

66

68

70

72

0,01 0,1 1

1-kHz pure tone

Duration (s)

τ = 127 msCritical duration = 381 ms

y = 71.2+10*log(1-exp(-x/0,127))R² = 0,93Lo

udne

ss le

vel (

phon

s)

From Boullet (2005)

PhD Thesis

duration



ANSI, S3.4-2005, « Procedure for the Computation of Loudness of Steady Sounds, », American National Standards Institute, New York (2005).

From Moore B. C. J. and Glasberg B. R., Acustica-Acta Acustica, 82, 335 (1996).

Moore B. C. J., Glasberg B. R., Baer T., J. Audio Eng. Soc., 45, 224 (1997).

ISO 532B, « Method for calculating loudness level », International Organisation for standardization (1975).

From Zwicker E., Acustica, 10, 304 (1960)

Zwicker E., J. Acoust. Soc. Am., 33, 248 (1961)

Loudness modelsStandards for steady sounds



Zwicker’s modelStationary sounds, free or diffuse field

Signal

FFT

Critical bands or 1/3 octave bands

Excitation

Specific loudness

Overall loudness

1 24

aD+ filtering

a0 filtering From free field to inner ear

From free field to diffuse field

Calculated using masking curves

Based on Stevens law

=Σ specific loudnesses(because of spectral loudness summation)



Moore and Glasberg’s model

Based on Zwicker’s model

Differences:

1 –Auditory filters shapes,

2 –Excitation pattern,

3 - a0 and aD+

Stationary sounds, free or diffuse field



Loudness models for non-stationary sounds

Zwicker E., “Procedure for calculating loudness of temporally variable sounds”, J. Acoust. Soc. Am., vol.62, n°3, 675-682, 1977.

Zwicker E. et Fastl H., “Psychoacoustics: Facts and models”, 2nd

Edit ion, Springer-Verlag, Berlin, 1999.

Glasberg B. R. and Moore B. C. J., “ A model of loudness applicable to time-varying sounds”, J. Audio Eng. Soc., 50, n°5, 331-342, 2002.



Current researches

Loudness of non-stationary sounds

Short duration signals

Long duration signals

Effect of context

Induce Loudness Reduction (Recalibration)

Loudness Constancy

Binaural Loudness Summation

Spectral loudness summation and duration




Short duration sounds

0

5

10

15

20

20 40 60 80 100 120

Short varied

Long varied

Leve

l Diff

eren

ce (L

shor

t-Llo

ng) i

n dB

Level of Short Tone in dB SPL

Temporal integration = Lshort – Llong

short and long signal at equal loudness

Temporal integration

depends on levelTemporal integration maximum for moderate levels

From Florentine et al., JASA 1996



From Florentine et al., JASA 1996



0

5

10

15

20

20 40 60 80 100 120

Short varied

Long varied

Leve

l Diff

eren

ce (L

shor

t-Llo

ng) i

n dB

Level of Short Tone in dB SPL

200-ms Equal-Ratio Loudness5-ms Equal-Ratio Loudness

Modified Power Function

Lshort – Llong =19dB

Lshort – Llong =14dB

Loudness functions are not linear





These loudness functions show features similar to the mechanicalinput/output measurement at the basilar membrane

Temporal integration of loudness does not depends on level



Loudness of non-stationary soundsEnvironmental short-duration

sounds

From Meunier et al., ICA 2001, Forum Acusticum 2002

Most studies:Rectangular envelope

Environmental sounds:Exponential envelope

N = kEaTb N: loudnessE: energyT: sound durationa, b: constants



N= kEaTb

Determination of a and b

a: loudness functions for environmental short duration sounds

b: loudness as a function of duration

Loudness of non-stationary soundsEnvironmental short-duration

sounds



- Loudness is constant when Energy increases as duration increases

- Loudness is constant when Energy is constant:

Equal energy rulefor duration less than the critical duration

loudness = constant if Energy = Intensity x duration = constant

- Loudness is constant when Energy decreases as duration increases


Environmental short duration sounds

Loudness as a function of durationThree relationships between loudness and duration were found in different studies :



dttpET 2

0

)((∫=Energy = = constant



Loudness as a function of duration

From Meunier et Rabau., Acoustics 08



0,1

1

1 10 100 1000

88 dB HL73 dB HL58 dB HL

60dB SL45dB SL

Estim

ated

loud

ness

(nor

mal

ized

)

Duration (ms)

bands of noiseF0=1 kHzΔf= 80 Hz

From Meunier et Rabau., Acoustics 08



Loudness as a function of duration

When Energy constant, Loudness varies as a power function of Signal Duration.

Exponent depends on the level of the signal.

Softer signals:Loudness constant when Energy decreases as duration increases

Louder signals:Loudness constant when Energy constant as duration increases



From Boullet et al. in preparation


Loudness Model for Impulsive

Sound (LMIS)

20

30

40

50

60

70

80

90

100so

n23

son2

2so

n02

son0

5_L3

son0

3so

n24

son0

9so

n15

son0

1so

n05_

L2so

n06

son0

4so

n13

son0

5_L1

son1

9so

n16

son2

0so

n08

son1

2so

n11

son1

7so

n18

son0

7so

n21

son1

4so

n10

Measured Loudness LevelLMIS

Loud

ness

Lev

el (p

hons

)

Sound n°




Long duration sounds

How does listeners judge overall loudness of time-varying sounds ?

Time (s)

Loudness N

N10

N5?

1 min




Long duration sounds

Kuwano and Namba (Psychol. Res., 1985) and Fastl (5th Oldenburg Symp. Psych. Acoutics, 1991):

Sound events proeminent in level strongly influence global loudness

Susini et al. (Acta Acustica, 2002):Recency effect: related to the temporal position of the highest contour peak

Global loudness: combination of highest levels, of their temporal position and their duration of emergence



John G. Neuhoff , Nature, 1998, 395, 123-124

Loudness of non-stationary soundsTemporal asymmetry

Loudness change: asymmetric

Asymmetry depends on:direction of change (increasing vs. decreasing)range of levels (high vs. low).

Loudness change of tones with linearly varying levels



John G. Neuhoff , Nature, 1998, 395, 123-124

Loudness of non-stationary soundsTemporal asymetry


Lannion, France, 10-12 September 200835Canévet et al., Acta Acustica, 2003

Loudness of non-stationary soundsTemporal asymetry

1

10

100

1 10 100

75->4545->7560->7575->60

Rat

io o

f est

imat

ions

(Lou

der

divi

ded

by s

ofte

r)

Sweep duration in seconds

A) - Pure tones at 1 kHz

(17 subjects) decreasingincreasingincreasingdecreasing



Effect of contextInduced Loudness Reduction (ILR)

A preceding higher-level tone (inducer) reduces the loudness of a lower-level tone (test tone)

Marks, J. Acoust. Soc. Am., 1996

A: 500-Hz tones relatively low SPLs and 2500-Hz tones high SPLs

B: reverse



Effect of contextInduced Loudness Reduction (ILR)

Amount of ILR depends on:Tone levelsFrequency separation between inducer and test toneDuration of inducer and test toneTime separation between inducer and test toneIndividual differences

Review in Epstein, J. Acoust. Soc. Am., Online, 2007



Effect of contextLoudness Constancy

Zahorik and Wightman, Nature neuroscience, 2001

Loudness Constancy

Loudness contant for fixed source power and variable

source distance

source power

source power

source power

Intensity changes at the ear may be due to• Source power changes• Source distance



Effect of contextBinaural Loudness Summation

(BLS)

From Marozeau et al. J. Acoust. Soc. Am., 2006

Binaural loudness = A x monaural loudnessA: from 1.3 to 2 depending on study



Effect of contextBLS as a function of stimulus and

listening conditionsStimuli• Monitored Live Voice (MLV) spondees

• Recorded spondees• Tones

Listening conditions• Earphone• Loudspeakers

From Florentine J. Acoust. Soc. Am., 2008

BLS for tones or recorded spondees > BLS for MLV

BLS for earphones > BLS for Loudspeakers

BLS in laboratory conditions > BLS out of the laboratory







Excitation

Masking curves



Physical parameters of the twelve synthesized noises

8 listeners

s o u n dn u m b e r

C e n t r a lf r e q u e n c y

( H z )

b a n d w i d t h( H z )

L e v e l( d B S P L )

1 4 0 0 5 0 4 02 4 0 0 4 2 0 4 03 4 0 0 5 0 6 04 4 0 0 4 2 0 6 05 1 4 2 0 1 2 0 3 06 1 4 2 0 1 0 0 0 3 07 1 4 2 0 1 2 0 5 08 1 4 2 0 1 0 0 0 5 09 3 0 0 0 2 4 0 6 0

1 0 3 0 0 0 2 0 4 0 6 01 1 3 0 0 0 2 4 0 7 01 2 3 0 0 0 2 0 4 0 7 0

From Meunier et al. InterNoise 2000

Experiment 1Loudness of synthesized noises



30405060708090

30 40 50 60 70 80 90

12

34

56

78

9 10

11 12

y = 7,5766 + 0,85313x R= 0,97504

Adj

uste

d lo

udne

ss

(pho

ns)

Calculated loudness (phons)

Moore's model (AES)

7

30405060708090

30 40 50 60 70 80 90

12

34

56

78

9 10

11 12

y = 1,893 + 0,95909x R= 0,98632

Adj

uste

d lo

udne

ss

(pho

ns)


Zwicker's model

30405060708090

30 40 50 60 70 80 90

12

3 4

56

8

9 10

11 12

y = 0,8725 + 0,9377x R= 0,99454

Adj

uste

d lo

udne

ss

(pho

ns)


Moore's model (Acustica)

7


Experiment 1 : synthesized noises



Twenty four enrironmental sound (steady over 1 s)Sound Abbreviation

Blowlamp BlowlampGuitare GuitareHarmonica HarmRumpled paper PaperComputer hard disk DiskTelephon in an Anecho•c Chamber Tel_ACTelephon in an office TelBicycle in an Anecho•c Chamber Bicy_AcBicycle BicyCar CarWoman voice Voice_WMan voice Voice_MFlute at 39 dB SPL Flute_39Flute at 54 dB SPL Flute_54Flute at 69 dB SPL Flute_69Flute at 84 dB SPL Flute_84Motorcycle at 28 dB SPL Moto_28Motorcycle at 43 dB SPL Moto_43Motorcycle at 58 dB SPL Moto_58Motorcycle at 73 dB SPL MotoDrilling at 35 dB SPL Drill_35Drilling at 50 dB SPL Drill_50Drilling at 65 dB SPL Drill_65Drilling at 80 dB SPL Drilling

24 listeners


Experiment 1Loudness of environmental noises



30405060708090

100

30 40 50 60 70 80 90 100

y = 15,611 + 0,77787x R= 0,98295

Adj

uste

d lo

udne

ss

(pho

ns)


Zwicker's model

405060708090

100

40 50 60 70 80 90 100

y = 3,6762 + 0,87937x R= 0,97893

Adj

uste

d lo

udne

ss

(pho

ns)


Moore's model (Acustica)

405060708090

100

40 50 60 70 80 90 100

y = 8,587 + 0,82236x R= 0,96235

Adj

uste

d lo

udne

ss

(pho

ns)


Moore's model (AES)


Experiment 2 : environmental noises

Documents

Loudness: Current Knowledge and Questions - ITU · International Telecommunication Lannion, France, 10-12 September 2008 Union Loudness: Current Knowledge and Questions Sabine Meunier,