=Paper=
{{Paper
|id=Vol-2068/wii6
|storemode=property
|title=A Smart Walking Navigation System based on Perceived Exertion
|pdfUrl=https://ceur-ws.org/Vol-2068/wii6.pdf
|volume=Vol-2068
|authors=Panote Siriaraya,Kodo Maeda,Yusuke Nakaoka,Yukiko Kawai,Shinsuke Nakajima
|dblpUrl=https://dblp.org/rec/conf/iui/SiriarayaMNKN18
}}
==A Smart Walking Navigation System based on Perceived Exertion==
https://ceur-ws.org/Vol-2068/wii6.pdf
A Smart Walking Navigation System
based on Perceived Exertion
Panote Siriaraya Kodo Maeda Yusuke Nakaoka
Kyoto Sangyo University Kyoto Sangyo University Kyoto Sangyo University
Kyoto, Japan Kyoto, Japan Kyoto, Japan
spanote@gmail.com g1445232@cc.kyoto-su.ac.jp g1444936@cc.kyoto-su.ac.jp
Yukiko Kawai Shinsuke Nakajima
Kyoto Sangyo University Kyoto Sangyo University
Kyoto, Japan Kyoto, Japan
kawai@cc.kyoto-su.ac.jp nakajima@cse.kyoto-su.ac.jp
ABSTRACT government policies and initiatives (such as [9]) have been put
In recent years, the growing trend of personal well-being has in place to help raise awareness about the importance of phys-
led an increasing number of people to engage in walking re- ical exercise. However, even though people are aware of the
lated activities to improve their physical health. As such, it importance of physical exercise, it is difficult for them to en-
is becoming increasingly important to develop walking nav- gage and continue in such activities for a prolonged period of
igation systems which are able to effectively support such time. As such, there is an emerging need for the development
activities. The conventional walking navigation systems found of walking navigation systems which are able to effectively
in modern smart phone devices usually recommends the short- support health promoting walking activities.
est route to a destination and are not aimed at facilitating
Despite walking navigation systems becoming more common-
health promoting walking activities. Therefore, in this paper,
place in modern smart phone devices, the conventional navi-
we propose a novel walking navigation system which takes
gation systems were not created with the goal of facilitating
into consideration the rate of perceived exertion (based on the
health related walking activities and generally recommends
walking distance, geographical elevation and heart rate data)
the shortest route to a selected destination. In this paper, we
and the avoidance of dangerous locations and the possibility
propose a smart walking navigation system which is aimed
of passing through highly rated scenic locations (based on
especially at supporting health related walking activities. To
social network data from twitter) when recommending a route.
create a walking route which is effective, enjoyable and safe
Finally, we describe the results of a feasibility experiment
for users, we describe how an appropriate perceived exertion
study carried out to test the routes in the system.
load could be determined based on information such as the
walking duration, geographical inclination and heart rate and
ACM Classification Keywords
describe how the system could help users avoid accident prone
H.5.1 Multimedia information Systems: Hypertext naviga- locations and suggest pleasurable walking routes which passes
tion and maps through scenic locations by using social network data from
twitter.
Author Keywords
Waking Navigation; Walking Support Systems; Geotagged The appropriate perceived exertion load refers to an exertion
tweet analysis; Danger Avoidance load which is appropriate for a user based on their physical
characteristics and their purpose of exercise. Overall, the walk-
INTRODUCTION ing load of each route is calculated by combining data from
The recent growing trend of personal well-being has led an the measurement of the level of inclination and distance within
increasing number of people to engage in walking related the route with heart rate data. To help avoid dangerous routes,
activities for a variety of health related purposes, such as to data from geo-tagged twitters are also obtained and analyzed.
reduce weight and maintain their physical fitness. Various For instances, various phrases which denote a dangerous con-
ditions, such as "it’s dangerous", "I’m afraid" or "It’s dark" are
collected and used to estimate the level of danger on a route.
In addition, to facilitate a more enjoyable walking experience,
data which could indicate the presence of appropriate resting
locations is also taken into consideration. This includes data
which describes locations with good panoramic views or sea-
sonal sightseeing locations (for cherry blossom and autumn
©2018. Copyright for the individual papers remains with the authors. leaf viewing etc.). Similar to how we obtained information
Copying permitted for private and academic purposes.
WII’18, March 11, 2018, Tokyo, Japan
based on the number of calories which users wish to lose [14].
The overall aim of the system is to provide continuous support
for users who wish to take daily walks for exercise, with the
system attempting to recommend different routes each day to
prevent boredom. To help reduce stress during walking related
activities, research carried out by Kitabayashi et al [6] used
biological data such as the user’s heart rate and data related
to the walking environment to analyze and estimate the level
of stress. The main aim was to provide a walking path which
is physically less stressful, especially for populations such as
older people. Overall, even though such systems are helpful
in recommending a route which is appropriate as a means
of exercise, there is still the problem of motivating users to
participate in the walking activity. To better motivate users,
we believe that it would also be advantageous to also consider
information from the environment within the walking route
as well (such as potential scenic locations within the walking
route).
Various systems have also been developed to help provide
Figure 1. The concept map of the smart navigation system users with a more pleasurable walking experience. For in-
stance, one study proposed a route recommendation system
which is not based on the shortest distance, but on the potential
about dangerous routes, the system collects data about ap- enjoyment quality within the route [11]. Abdallah et al [2] in
propriate resting locations from geo-tagged tweets. Highly the meanwhile, proposed a route recommendation system for
evaluated scenic locations are identified and considered in the those seeking to explore a city. The aim was to provide users
route recommendation. with recommendations of routes which contain interesting el-
Overall, this paper presents a smart walking navigation system ements (scenic views etc.) by analyzing geo-tagged photos
which recommends a route based on the rate of perceived exer- from photosharing sites such as Flickr. Other researchers have
tion and the possibility of passing through highly rated scenic also proposed a walking support system which allows users to
locations and avoiding dangerous locations. In this paper, we share scenic locations within a walking course with others[8].
discuss how such a navigation system could be implemented. Although such systems are helpful in providing a motivating
Afterwards, we describe a feasibility experiment carried out walking route for users, they generally do not include biolog-
to test the routes in the proposed system. ical data from users in their recommendation and therefore
tend not to take into account the physical characteristics of
RELATED RESEARCH
users (such as their physical ability or exertion) which could
also significantly influence the walking experience.
Route recommendation systems have been developed and used
for a variety of purposes such as in tourism and health care
(see [12] [14] [1] etc.). For those aimed at supporting walking A SMART WALKING NAVIGATION SYSTEM BASED ON
related activities, previous research tend to focus on issues THE RATE OF PERCEIVED EXERTION
related to health and safety (such as providing a walking route
Overview
which can effectively improve the health of a user during exer-
In this paper, we propose a Smart Walking Navigation Sys-
cise activities) and issues related to improving the underlining
walking experience (such as providing a walking route which tem which recommends a route based on the user and path
is aesthetically pleasurable). characteristics, taking into account the rate of perceived exer-
tion and the possibility of avoiding dangerous locations and
Examples of walking support and route recommendation sys- the ability to pass through highly rated scenic locations. An
tems which focus on health and safety include one system overview of the system is shown in Figure 1. Users first start
devised by Takaishi et al [13] which aimed to provide older by inputing information about their age, gender, resting heart
users with a personalized walking plan by combining data rate and desired walking duration. Based on this, the system
from a GPS and a heart rate monitoring device. This was would then determine an appropriate walking distance. Within
done to provide better walking instructions for older people this walking distance area, the system would obtain data about
looking to go walking as a way to exercise and improve their the inclination level, the heart rate of previous pedestrians, the
health. However, the adoption and execution of this "walk- presence of potentially dangerous locations (such as narrow
ing plan" is still reliant on the motivation of the users and paths, heavy traffic and train crossings) and locations which
the system does not automatically change or recommend the are scenic or popular with tourists. Based on this information,
route in real time. In terms of safety, Kim et al.[5] proposed several walking routes would then be recommend to the user.
a system which helps users avoid dangerous routes by using In the following section, we provide more details about the
data from geo-tagged tweets. Another study proposed a route data used in the recommendation of the routes and discuss the
recommendation system which recommends a walking route method which can be used to obtain the data.
The data and phrases such as "dangerous" and "scary" from geo-tagged
heart rate data tweets.
To help determine the appropriate exertion load, the user’s tar-
get heart rate is calculated [4]. The equation used to calculate Implementation of the walking route recommendation
the target heart rate is as follows: system
thri = (220 - agei - rhri ) * tc + rhri Estimating the walking speed based on age and gender
When using the system, the users would input how long they
In the equation, thr refers to the target heart rate , rhr refers to wish to walk (their desired walking duration). To be able to
the resting heart rate and tc refers to the target coefficient. accurately determine the length of the recommended route, it
is necessary to estimate the potential walking speed of each
Rating Perceived Exertion Intensity Heart rate user. Information about the age and gender of the users is used
20 Maximum exertion 100.0 200 to help determine the walking speed. The average walking
19 Extremely hard 92.9 speed of people when they walk around indeterminately could
18 85.8 180 be found at [16]. Based on this data, the appropriate walking
17 Very hard 78.6 distance (i.e. the desired length of the route) is calculated.
16 71.5 160
15 Hard 64.3 Obtaining map and elevation data
14 57.2 140 The open source software QGIS was used to obtain Open
13 Somewhat hard 50.0 street map (OSM) data [7] as well as data about the elevation,
12 42.9 120 latitude and longtitude.The data obtained in this study is cen-
11 Light 35.7 tered around the city of Kyoto, Japan at the latitude-longitude
10 28.6 100 coordinate of 6,662. Every OSM data point was given eleva-
9 Very light 21.4 tion, longitude and latitude data from QGIS. The length of
8 14.3 80 the road was calculated using differences in longitude/latitude.
7 Extremely light 7.1 The inclination between two points was calculated based on
6 No exertion 0.0 60 the difference in elevation.
Table 1. The relationship between the Rate of Perceived Exertion and
heart rate as described by the Japanese Institute of Health and Exercise Route recommendation through Dijkstra’s algorithm
Traditional walking navigation systems calculate the shortest
Table 1 shows the relationship between the Rate of Perceived route between the starting point and target destination using
Exertion and heart rate. In this study, the intensity level Dijkstra’s algorithm. In this study, in addition to the walking
most effective for dieting (which has been defined as between distance, we would also take into account the perceived rate of
"Light" and "Hard") which is used as the target coefficient in exertion, the presence of potentially dangerous areas as well
the equation has been set to around 40 to 60 percent. as the availability of scenic locations.
The level of inclination Initially, the proposed system randomly selects multiple candi-
Geospatial elevation data from the the Geospatial Information date sites based on the target walking distance. The inclination
Authority of Japan was used to calculate the level of inclination data and information about possible dangerous and scenic lo-
for a specific location [3]. The inclination data was acquired cations are obtained for the points on the shortest route to the
by using elevation data from a 5 meter grid interval. The open candidate sites (calculated using Dijkstra’s algorithm). Af-
source software QGIS [10] was used to create a road network terwards, the system recommends routes by considering the
containing the latitude, longitude and elevation information. perceived exertion load and the possibility of passing through
safe and scenic locations.
Information about scenic locations, word of mouth reviews and
For example, if a walking distance of 800 meters is required,
dangerous locations
multiple points from within the edge of the approximately 800
Information about scenic locations in the area could be ob- meter circular radius range are selected and the shortest route
tained from geo-tagged tweets combined with data from sight- from the starting spot to these points are calculated using the
seeing guide websites. For instance, tweets with phrases such Dijkstra’s algorithm. When the desired walking distance is
as "the cherry blossoms are beautiful" and "The night view reached, the points are then made into candidate sites. See
is pretty" could provide information about possible scenic Figure 2 for an example of the routes proposed by the system.
locations in the area. A key advantage of using geo-tagged
tweets is that it allows us to obtain information about not only After the candidate sites have been obtained, data about the
the popular tourist-based scenic locations in the area, but also inclination would be acquired for the points on the routes to
allows us to identify less well known ones as well. In addition, the candidate sites and used to investigate the physical exertion
using geo-tagged tweets allow the system to obtain scenic loca- load based on the targeted heart rate of the user. In the future,
tion information which is more context specific and up-to date the heart rate data would be used to help predict the level
(such as flower viewing spots which only appear during cer- of perceived exertion on different routes. Geo-tagged tweets
tain seasons or night viewing spots which are most beautiful were also obtained for the points on the route. In regards to the
during night time). Information about potentially dangerous tweets, in the current system, the number of tweets were used
routes is obtained in a similar manner, by identifying words as an indicator of the popularity of the location, regardless of
whether they were positive or negative in nature. In the future, level of inclination. In addition,the routes to the destinations
we would look into using sentiment analysis to separate the G1 and G3 were also selected as they were similar in terms of
negative locations from the positive ones. Overall, both data tweet numbers but had large difference in inclination change.
related to the inclination and the tweets would be displayed Participants wore a smart watch which measured their heart
for the routes to the candidate sites. rate as they walked through the route. In addition, partici-
pants were also asked to rate their experience while walking
To test the feasibility of the currently proposed system, an the route. They were asked to rate three factors 1) how easy
experiment study was carried out. The main aim was to de- they felt the route was 2) how enjoyable was the route and
termine whether the perceived exertion load of the user is 3) whether they would like to walk on the route again on a 7
influenced by the inclination within the route and whether point Likert scale.
geo-tagged tweets could be a good indicator of the enjoyment
within the route. The results are shown in table 3. As expected, the average
heart rate of participants increased substantially for routes
FEASIBILITY STUDY OF THE ROUTE RECOMMENDATION with high upper elevation (G1). In addition, participants felt
SYSTEM it was easier to walk on routes with a low combined upper
An experiment study was carried out to test the feasibility of elevation (G2 and G3) than those with a high combined upper
the proposed method. Figure 2, shows an example of 8 dif- elevation (G1 and G4). Thus, the results suggest that the level
ferent waking navigation routes recommended by the method of inclination change within a route could be a good indicator
proposed in the paper. The targeted walking distance (from of the perceived exertion of users on that route.
the start location "S" to the points G1 to G8) is around 800m. The influence of the number of tweets on user enjoyment and
motivation was less clear. There seems to be a trend in that
for routes with a high number of tweets, the reported level of
enjoyment and motivation seems to be higher than routes with
a lower number of tweets (i.e. G1 being higher than G2 and
G3 and G4 being higher than G3). Overall, the system could
be further refined by using methods such as sentiment analysis
to analyze the tweets in more detail, using a method similar to
[15] to incorporate the tweets to a route area.
Ave Ave Rating Ave Rating Ave Rating
Route
Heart rate Ease Enjoyment Motivation
G1 100.0 3.3 5.7 5.7
G2 88.3 6.0 5.0 4.3
G3 89.0 5.7 3.0 3.7
Figure 2. The 8 example recommended routes
G4 87.7 3.7 5.0 4.3
Table 3. The results of the feasibility experiment study (N=3)
Combined combined Number SUMMARY
Target Total
upper lower of In this paper, we have proposed a smart walking navigation
destination distance
elevation elevation tweets system which recommends a walking route that takes into
G1 795m +13m -3m 14 tweets account the perceived exertion load (based on the walking
G2 794m +3m -5m 9 tweets duration, heart rate and inclination etc.), the possibility of
G3 801m +1m -11m 8 tweets avoiding dangerous locations and the ability to pass through
G4 799m +9m -6m 43 tweets highly rated scenic and popular locations. The overall aim
G5 795m +11m -2m 12 tweets of the system is to provide a walking experience which is
G6 794m +4m -3m 26 tweets enjoyable, safe and beneficial to the physical health of the user.
G7 806m +1m -11m 25 tweets We have described how such a system could be implemented.
G8 802m +3m 0m 30 tweets In addition, we have provided the results of an experiment
Table 2. Information about the inclination of the recommended route study carried out to test the feasibility of the proposed method,
and the number of tweets showing how the routes suggested by the system could influ-
ence the level of perceived exertion for users. In the future, we
Table 2 shows the total distance, inclination and total number
aim to refine the proposed method by using sentiment analysis
of tweets from the start location to the target destinations
to analyze the tweets and determine the scenic and dangerous
points (G1 to G8).
locations within the route. In addition, we would also exam-
To test the feasibility of the recommendation method, three ine a more personalized approach towards recommending the
participants were asked to walk through the routes recom- appropriate walking distance for each user based on their phys-
mended by the system. Overall, the routes to the destinations ical abilities and health conditions (i.e. by using information
G2 and G4 were selected as they had a large difference in from smart watches which can collect health related data to
tweet numbers but were similar in terms of changes to the infer the physical ability of users etc.).
REFERENCES 10. QGIS. 2017. About QGIS. (2017). Retrieved Aug 29,
1. Keith Cheverst, Nigel Davies, Keith Mitchell, Adrian 2017 from http://qgis.org/ja/site/.
Friday, and Christos Efstratiou. 2000. Developing a
11. Daniele Quercia, Rossano Schifanella, and Luca Maria
context-aware electronic tourist guide: some issues and
Aiello. 2014. The shortest path to happiness:
experiences. In Proceedings of the SIGCHI conference on
Recommending beautiful, quiet, and happy routes in the
Human Factors in Computing Systems. ACM, 17–24.
city. In Proceedings of the 25th ACM conference on
2. Abdallah El Ali, Sicco NA Van Sas, and Frank Nack. Hypertext and social media. ACM, 116–125.
2013. Photographer paths: sequence alignment of
geotagged photos for exploration-based route planning. 12. Marta Rey-López, Ana Belén Barragáns-Martínez, Ana
In Proceedings of the 2013 conference on Computer Peleteiro, Fernando A Mikic-Fonte, and Juan C Burguillo.
supported cooperative work. ACM, 985–994. 2011. moreTourism: Mobile recommendations for
tourism. In Consumer Electronics (ICCE), 2011 IEEE
3. Geospatial Information Authority of Japan. 2017. International Conference on. IEEE, 347–348.
Geospatial Information. (2017). Retrieved Aug 29, 2017
from http://fgd.gsi.go.jp/download/menu.php. 13. Tetsuo TAKAISHI, Yoshie YAMADA, Tsutomu
TANAKA, Miyuki KANEWAKA, and Hisayo
4. Japanese Institute of Exercise and Health. 2017. Exercise YANAGISAWA. 2009. Recommendation of using a
useful for health promotion. (2017). Retrieved Aug 29, registering type GPS and a heart rate monitor for walking
2017 from http://www.jhei.net/exer/walking/wa02.html. instructions in the elderly. Nihon Koshu Eisei
Zasshi(JAPANESE JOURNAL OF PUBLIC HEALTH) 56,
5. Jaewoo Kim, Meeyoung Cha, and Thomas Sandholm.
3 (2009), 172–183. DOI:
2014. Socroutes: safe routes based on tweet sentiments.
http://dx.doi.org/10.11236/jph.56.3_172
In Proceedings of the 23rd International Conference on
World Wide Web. ACM, 179–182. 14. Yasufumi Takama, Wataru Sasaki, Takafumi Okumura,
6. Hiroki Kitabayashi, Tsuneo Ohnishi, Zhang Shun Pen, Chi-Chih Yu, Lieu-Hen Chen, and Hiroshi Ishikawa.
Yasuhito Asano, and Masatoshi Yoshikawa. 2014. Crowd 2015. Walking Route Recommendation System for
data analysis of walking environment and biomedical Taking a Walk as Health Promotion. In Web Intelligence
signals for route recommendation by stress prediction. and Intelligent Agent Technology (WI-IAT), 2015
IPSJ SIG Notes 2014, 5 (jul 2014), 1–6. IEEE/WIC/ACM International Conference on, Vol. 1.
http://ci.nii.ac.jp/naid/110009808961/en/
IEEE, 556–559.
7. Learn OSM. 2017. OSM data overview. (2017). Retrieved 15. Shoko Wakamiya, Hiroshi Kawasaki, Yukiko Kawai,
Aug 29, 2017 from Adam Jatowt, Eiji Aramaki, and Toyokazu Akiyama.
http://learnosm.org/ja/osm-data/data-overview/.
2016. Lets not stare at smartphones while walking:
memorable route recommendation by detecting effective
8. Takeru Muto, Kiichiro Sasaki, and Takami Yasuda. 2014. landmarks. In Proceedings of the 2016 ACM
Consideration of walking support service with International Joint Conference on Pervasive and
Smartphones. The 76th National Convention of IPSJ Ubiquitous Computing. ACM, 1136–1146.
2014, 1 (2014), 229–230.
16. Yokohama city sports medical center. 2017. Normal
9. National Physical Activity Plan. 2013. U.S. NATIONAL walking speed. (2017). Retrieved Aug 29, 2017 from
PHYSICAL ACTIVITY PLAN. (MAR 2013). Retrieved https://www.yspc.or.jp/ysmc/column/health-fitness/
Aug 29, 2017 from http://www.physicalactivityplan. walking-2.html.
org/docs/2016NPAP_Finalforwebsite.pdf .