Transcript Diapositive 1

Acoustic descriptors for dynamic
noise estimation close to traffic
signals
Arnaud Can, LICIT (ENTPE/INRETS)
Ludovic Leclercq, LICIT (ENTPE/INRETS)
Joël Lelong, LTE (INRETS)
Introduction
Introduction
Existing descriptors
New descriptors
Conclusion
 Descriptors set by legislation can hardly
capture urban traffic noise variations
 Temporal noise structure
urban soundscape quality
influences
 Dynamics noise models are now able to
assess LAeq,1s evolution
[Leclercq-2002] ; [De Coensel et al.-2005]
Need descriptors that reflect noise dynamics
Can / Leclercq / Lelong
2
Outline
Introduction

Conclusion
Show their weaknesses for noise dynamics assessment
New descriptors
characterization




New descriptors
Existing descriptors and urban traffic noise
dynamics


Existing descriptors
for
urban
traffic
noise
Focus on noise variations at a signal-cycle scale
Based on Mean noise pattern reconstitution
Evaluation of noise variations around this pattern
Conclusion
Can / Leclercq / Lelong
3
Experiment
Introduction

Existing descriptors
New descriptors
Conclusion
Traffic situation:

in front of a traffic signal
Cours Lafayette, Lyon (France)
Three lanes one way street
Street quite busy (1400veh/hour)

Measurement:


Acoustics: LAeq,1s evolution
Traffic: tgreen=50s, tred=40s, flow rates
Can / Leclercq / Lelong
4
Existing descriptors and urban
traffic noise dynamics
5
Existing descriptors
and urban traffic noise dynamics
Introduction

New descriptors
Conclusion
Limits of classical descriptors calculated over
long period scales (24h)


Existing descriptors
Unable to capture long-term or short-term noise variations ;
see proceedings
Limits of classical descriptors calculated over
short period scales
Can / Leclercq / Lelong
6
Limits of classical descriptors
calculated over short period scales
Existing descriptors
Introduction
New descriptors
Conclusion
+3dB
1%
LAeq is too sensitive to peaks of noise
Can / Leclercq / Lelong
7
Limits of classical descriptors
calculated over short period scales
Introduction

Existing descriptors
New descriptors
Conclusion
Rhythm of noise at traffic signal scale is not captured by
usual descriptors
Need specific descriptors
t = 90s  traffic
cycle duration
Can / Leclercq / Lelong
8
New descriptors for urban
traffic noise characterization
9
New descriptors
for urban traffic noise characterization
Introduction
Existing descriptors
New descriptors
Conclusion

Description of the mean noise pattern

statistical descriptors vs. mean noise pattern

LAeq,1s distribution

Noise variations around the mean noise pattern
Can / Leclercq / Lelong
10
Description of the mean noise pattern
Introduction

Existing descriptors
New descriptors
Conclusion
Traffic noise alternates between two levels
 How descriptors are related to these levels ?
 How estimate these two levels ?
Can / Leclercq / Lelong
11
Classical noise descriptors
and mean noise pattern
Introduction


Existing descriptors
New descriptors
Conclusion
Statistical descriptors are not related to mean noise
pattern
Lgreen and Lred do not reflect upper and lower levels
Can / Leclercq / Lelong
12
Study of noise distribution
Introduction
Existing descriptors
New descriptors
Conclusion
Two modes that correspond to each traffic signal phase
How characterize this distribution ?
Can / Leclercq / Lelong
13
Study of noise distribution
Introduction

New descriptors
Existing descriptors
Conclusion
bi-gaussian function:
Two modes

A1
f x =
exp
 1 2

 x  x1

 1 2





r²adj=0.9988
2
+
Difference
between
2
modes
A

2
2
Amplitude
of
modes
Standard
deviation of
modes

Green and red
phases


2
 x  x2 
Dynamics
 at

 the
 2 traffic
2 
signal scale

exp

Which one is
predominant

Noise
variations
within each
mode
Need to study variations around the mean noise pattern
Can / Leclercq / Lelong
14
Noise variations
around the mean noise pattern
Introduction
Existing descriptors
New descriptors
Conclusion
intensity of peaks
Periodicity and intensity of peaks:




NLmax>80
NL5>75
L5/cycle
Lmax/cycle
Can / Leclercq / Lelong
Rarefaction of calm periods:
NLmin>60
disappearance of calmperiods



NL95>65
L95/cycle
Lmin/cycle
15
Conclusion
Introduction

Existing descriptors
New descriptors
Conclusion
Usual descriptors fail to capture urban noise dynamics


When calculated over long period
When calculated over short period

Noise dynamics at traffic signals may be characterized by
the mean noise pattern

None usual descriptor is related to this pattern

Specific descriptors can be proposed:


Bi-gaussian fit  mean noise pattern
Traffic-scaled variations descriptors  variations around the
mean noise pattern
Can / Leclercq / Lelong
16
Further investigations
Introduction

Existing descriptors
New descriptors
Conclusion
Method allows differentiation between noise
situations:

Comparison between the point in front of a trafic cycle
and a point between two traffic cycle : proceedings

Generalization on more complicated scenarios
(calm point, close bus station, two ways street…)
Can / Leclercq / Lelong
17
Thank you
for your attention
18
Limits of classical descriptors
calculated over long period scales (24h)
Introduction
Can / Leclercq / Lelong
Existing descriptors
New descriptors
Conclusion
19
Limits of classical descriptors
calculated over long period scales (24h)
Introduction

Existing descriptors
New descriptors
Conclusion
LAeq and statistical descriptors 24h estimation vs
LAeq1s evolution


Can / Leclercq / Lelong
Unable to capture long-term noise
variations [Can-2007]
Characteristics of the time slot are not
reflected by descriptors
20