Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Attractive routes
Catherine Lavandier
(Pauline Delaitre – Pierre Aumond)
Sound pleasantness of pedestrian walks in
urban environment
Sound pleasantness of a soundscape in urban situations
2
Aim of different projects was to propose urban sound quality indicators
based on perceptual and acoustic data, collected with different methods
• 60 participants (about 30 locations at different periods)
• 100 measurements each year per participant
• Only 3409 relevant data
Cart_ASUR project (2012-2016):
• 13 perceptual data collected with mobile phones in Paris
• acoustic data collected through the mobile microphone (dB(A) 1s)
3
Sound pleasantness of a soundscape in urban situations
GRAFIC project (2013-2017):
• perceptual data collected on questionnaires in Paris
• acoustic data collected through one mobile sensor (1/3 Oct band-125ms)
• georeferenced data for localisation of the measurements
4
Perceptual models of sound pleasantness
Cart_ASUR (3409 individual situations) Sound pleasantness = 8.11 – 0.38 * Global loudness – 0.14 * Time ratio of traffic + 0.20 * Time ratio of voices + 0.15 * Time ratio of birds (R²aj=0.34 & r = 0.58 and R²aj=0.80 & r = 0.94 for the mean values)
GRAFIC (556 individual situations) Sound pleasantness = 9,70 – 0,47 * Global loudness – 0.21 * Time ratio of traffic + 0.12 * Time ratio of voices + 0.09 * Time ratio of birds (R²aj=0.42 and R²aj=0,90 for the mean values)
5
Perceptual variable
Acoustic indicators (mean values Cart_ASUR
Acoustic indicators (mean values GRAFIC)
Global loudness
LAeq (0,76) L50 (0,82) L90 (0,71)
NNEL>70 (0,75)
LAeq (0,73) L50 1Khz (0,85)
L90 (0,72) N50 (0,80)
Time ratio of traffic
LAeq (0,63) L50 (0,62)
MIL>70 (0,66)
LAeq (0,66) L50 (0,76)
L50 250Hz (0,83) Time ratio of voices
σ (-0,61) LA10–LA90 (-0,61)
TFSD 500Hz,1s (0,53) LA10–LA90 (NS)
Time ratio of birds
L5 (-0,55) L10 (-0,53)
L10 (-0,46) TFSD 4kHz,125ms (0,81)
Acoustic models of sound pleasantness
6
Sound Pleasantness = 16.92 – 0.15 * LA50 - 0,06 * (LA10-LA90) (R²aj=60%)
Harmonica = 0.2*(LA95-30)+0.25*(LAeq-LA95) = - 6 + 0,25 * LAeq – 0.05 * LA95
Sound Unpleasant = -SP+11= - 5.9 + 0,15 * LA50 +0.06 * LA10 – 0.06 * LA90
Sound Pleasant. = 16.70 – 0.24 * L50 1KHz + 10.9 * TFSD 500Hz + 17.1* TFSD 4KHz
(R²aj=85%)
Acoustic models of sound pleasantness
Cart_ASUR (mean values)
GRAFIC (mean values)
CENSE (Running project)
Sound Pleasant. = ft (L50 + new indicators for presence of sources extracted from (1/3 Oct band-125ms) in real time with deep learning networks)
7
Sound pleasantness of a pedestrian route
For each assessment location Additional questions about segment between the assessment locations (~1min): • perceived pleasant. of the route • strength and velocity of the change of the sound environ.
20
dB
At the middle and the end of the route (~15min) •perceived pleasantness of the half routes and total route
8
S[1-2] S[2-3] S[3-4] S[4-5] S[5-6] S[6-7] S[7-8] S[8-9] S[9-10]
∆SP -0,83 -3,26* -0,61 -0,24 2,05 0,43 1,12 0,72 -0,62
SC 3,5 8,7 5,3 5,4 6,3 3,2 6,7 4,6 8,1
S[10-11] S[11-12] S[12-13] S[13-14] S[14-15] S[15-16] S[16-17] S[17-18] S[18-19]
∆SP -1,54 1,35 0,62 1,33 1,63 -0,52 -2,84* 0,10 -1,62
SC 6,5 5,9 5,6 4,9 8,2 2,5 8,8 6,2 8,0
SC |∆SP| Recency effect
SC = perceived change over 1min
GP= Global Pleasantness of the route
∆SP= difference of route pleasantness between directions
Impact of the direction of the route
Global Pleasantness = 0.85 + 0.44 * Pmean + 0.45 * Pend (R²aj=83%)
9
Impact of the duration of the route
Global Pleasantness over 15-minute route is a linear function of the average of the Pleasantness of locations which were assessed during the route. (r= 0.8 R²=64%) No recency effect here, - Only 12 mean data - Long term memory ?
10
Pleasantness in a laboratory context
Continuous assessment of the sound pleasantness At the end of the route (3-min stimuli) • perceived pleasantness of the route
• Semi-anechoic room
• Immersive video projection
• Calibrated sound levels
• Professional sound interface
• High quality monitoring
loudspeakers
• Transaural audio restitution
11
Construction of the stimuli
Controlled audio sequences
30 participants
16 paths from 2 mixed sound
environments (park and boulevard)
Calibrated binaural microphones
Blurred fix image 20
20
Audiovisual sequences
30 participants
5 paths in both directions
Recordings (3 minutes)
Video – Mini camera
12
Audio controlled sequences only, 3-min long
Global Pleasantness = -1.4 + 0.96 * Pmean + 0.22 * Pend
(R²aj=95%) (R²aj=75% if Pmean only)
13
Audio visual sequences, 3 minute long
Global Pleasantness over 3-minute route is a linear function of the average of the Pleasantness of locations which were assessed during the route. (Raj²=75%). No recency effect with complex sequences.
Impact of the vision/complexity on the change in sound pleasantness
14
Change of SP Saliency 21.5%
Δ% = 27.2% Audio only
Δ% = 14.7% Audio-Visual
Saliency Change of SP 48.7%
Saliency Change of SP 47.5%
Change of SP Saliency 32.8%
Filipan K. et al., "Auditory sensory saliency as a better predictor of change than sound amplitude in pleasantness assessment of reproduced urban soundscapes", Building and Environment, Vol 148, pp. 730-741, (2018).
Summary
15
Sound quality of an attractive route
1-min walk (In Situ)
3-min experiment (Laboratory – Audio only)
3 min. (Laboratory – Audio Visual)
More than a 15-min walk (In Situ)
Recency effect has been shown but only for short or very controlled paths.
GP = 0.85 + 0.44 * Pmean + 0.45 * Pend (R²aj=83%)
GP = -1.4 + 0.96 * Pmean + 0.22 * Pend (R²aj=95%)
GP = Pmean (R²aj=64%-75%)
Good approximation with the average of the pleasantness of the successive locations.
Next steps – Sound Pleasantness of a route
16
Next steps – CENSE Smart City Project
17
• Classical acoustic indicators out of 120 sensors in a 1 Km² area in Lorient
• Automatic extraction of time of presence for different types of sound sources
• Automatic calculation
of Sound Pleasantness
Sound Pleasant. = ft (L50 + new indicators for presence of source extracted from (1/3 Oct band-125ms) in real time with deep learning networks)
Next steps – Interactive maps
18
• Interpolation of Sound Pleasantness between sensors (about 50m between each sensor).
• Automatic extraction of
time of presence for • Natural • Human • Traffic sound sources, shown with dedicated signs on the map.
Global Sound Pleasantness of a route ? Improve the models !
Attractive routes
Catherine Lavandier
(Pauline Delaitre – Pierre Aumond)
Thank you very much for your attention