Motion Period Identification for Real-Time Application with an Accelerometer
碩士 === 國立成功大學 === 工程科學系碩博士班 === 100 === Inertial measurement units (IMUs) have been widely applied in the human-machine interference. Their main application is in motion recognition, e.g. handwriting recognition, whereas such short-time motion recognition is almost realized by off-line algorith...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2012
|
Online Access: | http://ndltd.ncl.edu.tw/handle/75000477723373684933 |
id |
ndltd-TW-100NCKU5028072 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-100NCKU50280722015-10-13T21:33:37Z http://ndltd.ncl.edu.tw/handle/75000477723373684933 Motion Period Identification for Real-Time Application with an Accelerometer 動作時區辨識於加速度計之即時性應用 Yi-MingLiu 劉宜明 碩士 國立成功大學 工程科學系碩博士班 100 Inertial measurement units (IMUs) have been widely applied in the human-machine interference. Their main application is in motion recognition, e.g. handwriting recognition, whereas such short-time motion recognition is almost realized by off-line algorithms. It is crucial to accurately identify the motion period while using IMUs. The traditional algorithm is implemented with the received signals from IMUs whose high frequency noise is removed by a moving average filter. After doing experiments and analysis, we found, that actual moving average filtering is not necessarily required. In this thesis, we propose a new real-time algorithm of motion period identification that is able to characterize the Autocorrelation History (AH) of the received raw signals. This new algorithm has three AH thresholds for detecting the beginning, middle, ending of an action. In comparison with to other algorithms, these three thresholds can be set easily and flexibly since their ranges are fairly large and different. Our experimental results show that this new real-time algorithm can make the ratio of the integration error below 10%. Finally, we use an autoregressive (AR) model to model the action signals. This AR model can make the integration result match that of integration error accumulation, which is very helpful while doing error compensation. Jer-Nan Juang 莊哲男 2012 學位論文 ; thesis 56 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立成功大學 === 工程科學系碩博士班 === 100 === Inertial measurement units (IMUs) have been widely applied in the human-machine interference. Their main application is in motion recognition, e.g. handwriting recognition, whereas such short-time motion recognition is almost realized by off-line algorithms. It is crucial to accurately identify the motion period while using IMUs. The traditional algorithm is implemented with the received signals from IMUs whose high frequency noise is removed
by a moving average filter. After doing experiments and analysis, we found, that actual moving average filtering is not necessarily required.
In this thesis, we propose a new real-time algorithm of motion period identification that is able to characterize the Autocorrelation History (AH) of the received raw signals. This new algorithm has three AH thresholds for detecting the beginning, middle, ending of an action. In comparison with to other algorithms, these three thresholds can be set easily and flexibly since their ranges are fairly large and different. Our experimental results show that this new real-time algorithm can make the ratio of the integration error below 10%. Finally, we use an autoregressive (AR) model to model the action signals. This AR model can make the integration result match that of integration error accumulation, which is very helpful while doing error compensation.
|
author2 |
Jer-Nan Juang |
author_facet |
Jer-Nan Juang Yi-MingLiu 劉宜明 |
author |
Yi-MingLiu 劉宜明 |
spellingShingle |
Yi-MingLiu 劉宜明 Motion Period Identification for Real-Time Application with an Accelerometer |
author_sort |
Yi-MingLiu |
title |
Motion Period Identification for Real-Time Application with an Accelerometer |
title_short |
Motion Period Identification for Real-Time Application with an Accelerometer |
title_full |
Motion Period Identification for Real-Time Application with an Accelerometer |
title_fullStr |
Motion Period Identification for Real-Time Application with an Accelerometer |
title_full_unstemmed |
Motion Period Identification for Real-Time Application with an Accelerometer |
title_sort |
motion period identification for real-time application with an accelerometer |
publishDate |
2012 |
url |
http://ndltd.ncl.edu.tw/handle/75000477723373684933 |
work_keys_str_mv |
AT yimingliu motionperiodidentificationforrealtimeapplicationwithanaccelerometer AT liúyímíng motionperiodidentificationforrealtimeapplicationwithanaccelerometer AT yimingliu dòngzuòshíqūbiànshíyújiāsùdùjìzhījíshíxìngyīngyòng AT liúyímíng dòngzuòshíqūbiànshíyújiāsùdùjìzhījíshíxìngyīngyòng |
_version_ |
1718066561737031680 |