Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Identifying Different Motions Using Statistical Methods

Liu, Owen LU (2019) In Bachelor's Theses in Mathematical Sciences MASK01 20182
Mathematical Statistics
Abstract
This bachelor thesis aims to explore how well one can classify different types of motions using only data gathered from a mobile phones
gyroscope and accelerometer. The methods includes extracting features from the covariance estimations of each signal and taking timefrequency transforms of the data and classify the transforms with a
convolutional neural network. The results are positive and shows there
are ways to design algorithms that classify the motions correctly.
Popular Abstract
To many of us, the mobile phone has become an integrated part of our life. It helps us keep contact with friends but it also helps us during our every day life with various tasks, such as a calendar. While I did not construct an app, I explored to possibilities let the mobile phone aid us in an exercise session. With the help of the gyroscope and accelerometer - which keeps track of the phones rotation and acceleration respectively - to together keep track of how you move around, and use that information to track how many repeats of each motion you do.

The data that has been used in this study comes from 5 individuals, including myself, who in total produced 12 datasets. For each dataset, the individual performing the motions got the... (More)
To many of us, the mobile phone has become an integrated part of our life. It helps us keep contact with friends but it also helps us during our every day life with various tasks, such as a calendar. While I did not construct an app, I explored to possibilities let the mobile phone aid us in an exercise session. With the help of the gyroscope and accelerometer - which keeps track of the phones rotation and acceleration respectively - to together keep track of how you move around, and use that information to track how many repeats of each motion you do.

The data that has been used in this study comes from 5 individuals, including myself, who in total produced 12 datasets. For each dataset, the individual performing the motions got the same phone strapped to their upper arm, and was told how many repeats of what motion to do. For example, one could be told to do 10 push-ups, take a small break, then do 10 more push-ups, then finish.

Two different approaches when analyzing the data was done, one approach consisted of exploiting statistical properties of the data, and the other approach included time-frequency analysis and neural networks.

From the results, one can conclude that it definitely is possible to classify these motions well with both the approaches, and thus it feels plausible that an app to keep track of your training can be constructed. (Less)
Please use this url to cite or link to this publication:
author
Liu, Owen LU
supervisor
organization
course
MASK01 20182
year
type
M2 - Bachelor Degree
subject
publication/series
Bachelor's Theses in Mathematical Sciences
report number
LUNFMS-4039-2019
ISSN
1654-6229
other publication id
2019:K27
language
English
id
8996900
date added to LUP
2019-11-26 13:07:08
date last changed
2019-11-26 13:07:08
@misc{8996900,
  abstract     = {{This bachelor thesis aims to explore how well one can classify different types of motions using only data gathered from a mobile phones
gyroscope and accelerometer. The methods includes extracting features from the covariance estimations of each signal and taking timefrequency transforms of the data and classify the transforms with a
convolutional neural network. The results are positive and shows there
are ways to design algorithms that classify the motions correctly.}},
  author       = {{Liu, Owen}},
  issn         = {{1654-6229}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Bachelor's Theses in Mathematical Sciences}},
  title        = {{Identifying Different Motions Using Statistical Methods}},
  year         = {{2019}},
}