Distance estimations by non-linear step length models using accelerometer sensor data from waist-worn smart-phones
(2018) FMS820 20181Mathematical Statistics
- Abstract
- This thesis examines the possibility of modelling trajectory and estimating distance during a walk using
only a smart phone attached to the waist. Due to GPS not functioning indoors, a navigation system
without the need of external communication could be the perfect complement for closed spaces. This
thesis considers such a system, where the smart phone contains accelerometer and gyroscope sensors
for retrieving acceleration and angular rate measurements. Because of measurements errors, there is
no straightforward way to estimate the trajectory since double-integration creates a drift. There are
different ways to get around this problem, by using zero-velocity updates to reset the system or by
using biomechanical models of the human... (More) - This thesis examines the possibility of modelling trajectory and estimating distance during a walk using
only a smart phone attached to the waist. Due to GPS not functioning indoors, a navigation system
without the need of external communication could be the perfect complement for closed spaces. This
thesis considers such a system, where the smart phone contains accelerometer and gyroscope sensors
for retrieving acceleration and angular rate measurements. Because of measurements errors, there is
no straightforward way to estimate the trajectory since double-integration creates a drift. There are
different ways to get around this problem, by using zero-velocity updates to reset the system or by
using biomechanical models of the human walking patterns. The best approach explored in this thesis
is the latter, using non-linear models for estimating step lengths. The step models tend to under- or
overestimate and therefore is a scalable constant introduced to handle future distance estimations better.
The scalable constants are formed during several tests, using the least-squares method for minimizing
the total walked distance errors. The completed models are verified on a validation data set, confirming
the method being decent for distance estimation. In conclusion, the distance estimation works properly
good for implementation, but since the trajectory can not be reproduced from the non—linear step
models, there is a long way left before really contesting the GPS. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8950725
- author
- Drorsén, Benjamin and Strandberg, Daniel
- supervisor
- organization
- course
- FMS820 20181
- year
- 2018
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 8950725
- date added to LUP
- 2018-06-18 11:56:43
- date last changed
- 2018-06-18 11:56:43
@misc{8950725, abstract = {{This thesis examines the possibility of modelling trajectory and estimating distance during a walk using only a smart phone attached to the waist. Due to GPS not functioning indoors, a navigation system without the need of external communication could be the perfect complement for closed spaces. This thesis considers such a system, where the smart phone contains accelerometer and gyroscope sensors for retrieving acceleration and angular rate measurements. Because of measurements errors, there is no straightforward way to estimate the trajectory since double-integration creates a drift. There are different ways to get around this problem, by using zero-velocity updates to reset the system or by using biomechanical models of the human walking patterns. The best approach explored in this thesis is the latter, using non-linear models for estimating step lengths. The step models tend to under- or overestimate and therefore is a scalable constant introduced to handle future distance estimations better. The scalable constants are formed during several tests, using the least-squares method for minimizing the total walked distance errors. The completed models are verified on a validation data set, confirming the method being decent for distance estimation. In conclusion, the distance estimation works properly good for implementation, but since the trajectory can not be reproduced from the non—linear step models, there is a long way left before really contesting the GPS.}}, author = {{Drorsén, Benjamin and Strandberg, Daniel}}, language = {{eng}}, note = {{Student Paper}}, title = {{Distance estimations by non-linear step length models using accelerometer sensor data from waist-worn smart-phones}}, year = {{2018}}, }