Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote

碩士 === 長庚大學 === 資訊工程學系 === 99 === In this paper, we use the Nintendo Wiimote tri-axial accelerometer as an input device to make a gesture recognition system when using the Hidden Markov Model (HMM) as the recognition algorithm. We use a set of basic movements called “Fundamental Motions” as the synt...

Full description

Bibliographic Details
Main Authors: Wei Cheng Chen, 陳韋誠
Other Authors: R. Y. Lyu
Format: Others
Published: 2011
Online Access:http://ndltd.ncl.edu.tw/handle/18269101288733000450
id ndltd-TW-099CGU05392004
record_format oai_dc
spelling ndltd-TW-099CGU053920042015-10-13T20:27:49Z http://ndltd.ncl.edu.tw/handle/18269101288733000450 Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote 基於HMM的基礎動作合成之手勢辨識 及其於Wiimote上的實做 Wei Cheng Chen 陳韋誠 碩士 長庚大學 資訊工程學系 99 In this paper, we use the Nintendo Wiimote tri-axial accelerometer as an input device to make a gesture recognition system when using the Hidden Markov Model (HMM) as the recognition algorithm. We use a set of basic movements called “Fundamental Motions” as the synthesis of all the other complex motions. These Fundamental Motions are used as HMM modeling units. In our preliminary study, we use Arabic numerals '0 ' to '9' as the first recognition task. We analyze this task and find a set of 16 motions appropriate to be used as HMM modeling units. The second recognition task is Arabic numerals '10 ' to '99', we also use fundamental motion as main concept, but adding connection signal to represent the voice between models. We found the use of connection signal can increase the recognition rate about 30%. By using appropriate feature extraction and HMM topology, a HMM-Viterbi searching algorithm can achieve near 98% accuracy and 62.26% in average for making ten numbers in a set continuous gesture. R. Y. Lyu 呂仁園 2011 學位論文 ; thesis 90
collection NDLTD
format Others
sources NDLTD
description 碩士 === 長庚大學 === 資訊工程學系 === 99 === In this paper, we use the Nintendo Wiimote tri-axial accelerometer as an input device to make a gesture recognition system when using the Hidden Markov Model (HMM) as the recognition algorithm. We use a set of basic movements called “Fundamental Motions” as the synthesis of all the other complex motions. These Fundamental Motions are used as HMM modeling units. In our preliminary study, we use Arabic numerals '0 ' to '9' as the first recognition task. We analyze this task and find a set of 16 motions appropriate to be used as HMM modeling units. The second recognition task is Arabic numerals '10 ' to '99', we also use fundamental motion as main concept, but adding connection signal to represent the voice between models. We found the use of connection signal can increase the recognition rate about 30%. By using appropriate feature extraction and HMM topology, a HMM-Viterbi searching algorithm can achieve near 98% accuracy and 62.26% in average for making ten numbers in a set continuous gesture.
author2 R. Y. Lyu
author_facet R. Y. Lyu
Wei Cheng Chen
陳韋誠
author Wei Cheng Chen
陳韋誠
spellingShingle Wei Cheng Chen
陳韋誠
Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
author_sort Wei Cheng Chen
title Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
title_short Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
title_full Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
title_fullStr Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
title_full_unstemmed Gesture Recognition using HMM-based Fundamental Motion Synthesis with Implementation on a Wiimote
title_sort gesture recognition using hmm-based fundamental motion synthesis with implementation on a wiimote
publishDate 2011
url http://ndltd.ncl.edu.tw/handle/18269101288733000450
work_keys_str_mv AT weichengchen gesturerecognitionusinghmmbasedfundamentalmotionsynthesiswithimplementationonawiimote
AT chénwéichéng gesturerecognitionusinghmmbasedfundamentalmotionsynthesiswithimplementationonawiimote
AT weichengchen jīyúhmmdejīchǔdòngzuòhéchéngzhīshǒushìbiànshíjíqíyúwiimoteshàngdeshízuò
AT chénwéichéng jīyúhmmdejīchǔdòngzuòhéchéngzhīshǒushìbiànshíjíqíyúwiimoteshàngdeshízuò
_version_ 1718047127102291968