@techreport{oai:ipsj.ixsq.nii.ac.jp:00218013,
 author = {山本, 正明 and 中川, 弘充 and 松本, 茂紀 and 栗山, 裕之 and Masaaki, Yamamoto and Hiromitsu, Nakagawa and Shigenori, Matsumoto and Hiroyuki, Kuriyama},
 issue = {26},
 month = {May},
 note = {近年,COVID-19 (新型コロナウィルス感染症) の感染拡大に伴い,感染拡大防止と業務継続の両立に向けてオフィスワーカーの勤務場所が多様化した.そこで,在宅勤務であるかどうか,出社勤務であればどの座席を利用したかを位置推定する技術に注目が集まっている.これらの状況を鑑みて,スマートフォンを活用した機械学習での位置推定システムを検討している.しかし,機械学習に基づく位置推定システムは,スマートフォンの機種と保持状態に起因して位置推定を誤るという問題がある.そこで,スマートフォンの機種と保持状態に合わせて位置推定モデルを更新する再学習型位置推定システムを構築し,従業員 187 名を対象とした約 3 カ月間の実証実験を実施した.そして,再学習型位置推定システムは,スマートフォンの機種と保持状態に起因した位置推定誤差を 4.45m から 2.16m まで低減可能であることを明らかにした., Covid-19 is continuing to spread around the world. Office workers therefore work in several sites to balance work and prevetation of its. A positioning system for estimating worker’s positions is attracting considerable attention. Given this background, we propose a machine-learning based positioning system. In this system, a positioning model is generated from the training data measured by smartphones. The model estimates a worker’s position from the test data. When the training data and the test data were measured by using different types of smartphones, a positioning error increased. In addition, a different holding state of smartphone increased the error, too. To reduce the error, we developed a positioning system to update the training data according to the holding states and the types of smartphones. We conducted an experiment for three month with 187 employees to evaluate RMSE (Root Mean Square Error) of the positioning system. When using different holding states and the type of smartphone, RMSE of the system was reduced to 2.16m from 4.45m.},
 title = {電波強度に基づく再学習型位置推定システムの実証実験},
 year = {2022}
}