Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment...
Transcript of Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment...
![Page 1: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/1.jpg)
Human Perception of the Urban EnvironmentA Machine Learning Approach
Varun Ojha
![Page 2: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/2.jpg)
About the study
Human perception
Urban Environment
2
Dynamic environmental conditions such as noise, temperature,
illuminance, field of view, walkable area and traffic speed have the
potential to influence an individual’s perception.
Skin conductance responses of participants were measured with an E4
wearable device and processed to detect arousal levels as an indicator of
perception.
Wiedikon | Zürich | Switzerland
![Page 3: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/3.jpg)
Some facts
3
Diversity in days (participants walked on different days)
Diversity in experiment day’s time slot
Diversity in participant’s demographic profile
Uniformity in study location
Uniformity in season (month of April)
Data samples have the record of time and place
Human perception feature (physiological response) and urban
environment features are spatial-temporal data
Physiological responses are time-series data, thus need
special treatment.
Properties of dataset
Participant’s data
PN
P1 P2
P3 P4
P5 …
For 𝑖=1 to N, N = 30
Eve
nts
(tim
e a
nd
sp
ace
)
Features
P𝒊
![Page 4: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/4.jpg)
4Data cleaning
Examples of accepted physiological data for processing
Examples of rejected physiological data
![Page 5: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/5.jpg)
5Data smoothing and filtering
Original physiological data → Smooth physiological dataStationary Wavelet Transform
![Page 6: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/6.jpg)
6Data quantification
![Page 7: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/7.jpg)
7Arousal (nSCR) level detection
Ledalab for EDA signal analysisTypical signature of an Skin Conductance Response (SCR)
Source: Braithwaite, J. J., Watson, D. G., Jones, R., & Rowe, M. (2013). A guide for analysing electrodermal
activity (EDA) & skin conductance responses (SCRs) for psychological experiments. Psychophysiology, 49, 1017-
1034.
![Page 8: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/8.jpg)
Fusion
Participant’s data fusion
8
PN
P1
P2
P3
P4
P5
Features
Quantification
Eve
nts
(tim
e a
nd
sp
ace
)S
am
ple
s
Pi
<event - response>
For 𝑖=1 to N, N = 20
Attributes of the complied dataset
![Page 9: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/9.jpg)
Data labeling 9
Sense of physiological response (nSCR) labeling
Class 0 : A samples' physiological response value = 0
Normal physiological state
Class 1 : A samples' physiological response value > 0
Arousal physiological state
Box plot of nSCR values across all participants
![Page 10: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/10.jpg)
Sensitivity analysis
10
![Page 11: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/11.jpg)
Inference 11
![Page 12: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/12.jpg)
Feature’s importance
Environment measures:
S – Sound
D – Dust
T – Temperature
R – Humidity
I – Illuminance
Field of view measures:
A – Area
P – Perimeter
C – Compactness
Optimum set
![Page 13: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/13.jpg)
13Participants experience pattern
Physiological responsesClusters of participant's experience
SOM – 2D Map Label Map
Feature Map
![Page 14: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/14.jpg)
Fusion of data for Geo referencing
Participants data arrangement
14
PNP1P2P3P4 P5Features
Quantification
Eve
nts
(tim
e a
nd
sp
ace
)
Pi
Sa
mp
les
<event - response>
Average across all participants
For 𝑖=1 to N, N = 20
![Page 15: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/15.jpg)
15Geo-referenced average arousal Walkable spaceTraffic speed
Building construction year Façade color
![Page 16: Varun Ojha - ETH Z · 2018-04-20 · Varun Ojha. About the study Human perception Urban Environment 2 Dynamic environmental conditions such as noise, temperature, illuminance, field](https://reader031.fdocuments.us/reader031/viewer/2022040615/5f0f06057e708231d4421b80/html5/thumbnails/16.jpg)
16Conclusions
1. Conducted a real-life study to understand how human perceive their urban environment.
2. Perform fusion of information from multiple sensors responsible for recording environmental features and human physiological respo0nses.
3. Machine learning analysis affirmed that participants physiological responses were sensitive to slightest change in urban environment.
4. Machine learning analysis discovered that all the participants experienced similar environmental conditions, responded in a similar physiological arousal state.
5. Geo-referencing of participants physiological state enabled us to study further what was relation between participants physiological responses are dynamic urban environment such as traffic speed.