• ๋Œ€ํ•œ์ „๊ธฐํ•™ํšŒ
Mobile QR Code QR CODE : The Transactions of the Korean Institute of Electrical Engineers
  • COPE
  • kcse
  • ํ•œ๊ตญ๊ณผํ•™๊ธฐ์ˆ ๋‹จ์ฒด์ด์—ฐํ•ฉํšŒ
  • ํ•œ๊ตญํ•™์ˆ ์ง€์ธ์šฉ์ƒ‰์ธ
  • Scopus
  • crossref
  • orcid

  1. (Dept. of Electrical Engineering, Incheon National University, Korea.)
  2. (Dept. of Embedded System Engineering, Incheon National University, Korea.)



Healthcare, Wearable device, IMU, Motion classification, Transfer learning

1. ์„œ ๋ก 

์ตœ๊ทผ IoT ์‹œ๋Œ€์˜ ํ•ต์‹ฌ ๊ธฐ์ˆ  ์ค‘ ํ•˜๋‚˜์ธ ์›จ์–ด๋Ÿฌ๋ธ” ๊ธฐ๊ธฐ๊ฐ€ ํฐ ๊ด€์‹ฌ์„ ๋ฐ›์œผ๋ฉฐ ๋ฏธ๋ž˜ ์‚ฐ์—…์˜ ๊ธฐ๋Œ€์ฃผ๋กœ ๊ผฝํžˆ๊ณ  ์žˆ๋‹ค(1). ์›จ์–ด๋Ÿฌ๋ธ” ๊ธฐ๊ธฐ๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ์‹ ์ฒด ๋™์ž‘ ์ค‘์— ์ฐฉ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ๋งŒ๋“ค์–ด์ง„ ์ „์ž๊ธฐ๊ธฐ์ด๋‹ค. ์‹ ์ฒด์™€์˜ ์†Œํ†ต์ด ๊ฐ€๋Šฅํ•œ ์›จ์–ด๋Ÿฌ๋ธ” ๊ธฐ๊ธฐ๋Š” ์ตœ๊ทผ ์†Œํ˜•ํ™” ๋ฐ ๋ฐฐํ„ฐ๋ฆฌ ์„ฑ๋Šฅ์˜ ํ–ฅ์ƒ, ๋‹ค์–‘ํ•œ ๊ธฐ๋Šฅ์˜ ๊ฐœ๋ฐœ ๋“ฑ์œผ๋กœ ์ˆ˜์š”๊ฐ€ ๋”์šฑ ์ฆ๊ฐ€ํ•˜๊ณ  ์žˆ๋‹ค. ํŠนํžˆ COVID-19 ์œ ํ–‰ ์ดํ›„ ์ง‘๋‹จ ์ฒด์œกํ™œ๋™์— ๋Œ€ํ•œ ๊ทœ์ œ ๋ฐ ๊ฐ์—ผ ์šฐ๋ ค์— ๋”ฐ๋ผ ๊ฐœ์ธ์˜ ๊ฑด๊ฐ•๊ด€๋ฆฌ์— ํ™ˆ ํŠธ๋ ˆ์ด๋‹ ๋ฌธํ™”๊ฐ€ ์„ฑํ–‰ํ•˜๋ฉด์„œ ํ—ฌ์Šค์ผ€์–ด ๋ฐ ํ”ผํŠธ๋‹ˆ์Šค ์›จ์–ด๋Ÿฌ๋ธ” ๊ธฐ๊ธฐ์— ๋” ๋งŽ์€ ๊ด€์‹ฌ์ด ๋‚˜ํƒ€๋‚˜๊ณ  ์žˆ๋‹ค(2-3).

์›จ์–ด๋Ÿฌ๋ธ” ๊ธฐ๊ธฐ์—๋Š” ์ฐฉ์šฉ ํ˜•ํƒœ์— ๋”ฐ๋ผ ํฌ๊ฒŒ ํœด๋Œ€ํ•˜๋Š” ํ˜•ํƒœ์˜ ์ œํ’ˆ ๋ฐ ์•ก์„ธ์„œ๋ฆฌ์™€ ๊ฐ™์€ ์•ก์„ธ์„œ๋ฆฌํ˜•, ์˜๋ฅ˜ ํ˜•ํƒœ์ธ ์˜๋ฅ˜ ์ผ์ฒดํ˜•, ์‹ ์ฒด์— ๋ถ€์ฐฉํ•  ์ˆ˜ ์žˆ๋Š” ํ˜•ํƒœ์˜ ์‹ ์ฒด ๋ถ€์ฐฉํ˜•, ์‹ ์ฒด์— ์ง์ ‘ ์ด์‹ํ•˜๊ฑฐ๋‚˜ ๋ณต์šฉํ•˜๋Š” ํ˜•ํƒœ์˜ ์ƒ์ฒด ์ด์‹ํ˜•์œผ๋กœ ๋ถ„๋ฅ˜๋œ๋‹ค(4). ์ดˆ๊ธฐ์—๋Š” ์†๋ชฉ์— ์ฐฉ์šฉํ•˜๋Š” ์•ก์„ธ์„œ๋ฆฌํ˜•์ด ์ฃผ๋กœ ์‚ฌ์šฉ๋˜์—ˆ์œผ๋‚˜ ํ˜„์žฌ๋Š” ๋ณด๋‹ค ๋งŽ์€ ์‚ฌ์šฉ์ž์˜ ์ƒ์ฒด ๋ฐ์ดํ„ฐ๋ฅผ ์ธก์ •ํ•  ์ˆ˜ ์žˆ๋Š” ์˜๋ฅ˜ ์ผ์ฒดํ˜•์˜ ๋””๋ฐ”์ด์Šค๋„ ํ™œ๋ฐœํžˆ ๊ฐœ๋ฐœ๋˜๊ณ  ์žˆ๋‹ค(5).

์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ํ†ตํ•ด ๊ตฌํ˜„ํ•  ์ˆ˜ ์žˆ๋Š” ๋Œ€ํ‘œ์ ์ธ ๊ธฐ์ˆ ๋กœ ๋ชจ์…˜์บก์ณ ๊ธฐ์ˆ ๊ณผ ๋™์ž‘๋ถ„์„ ๊ธฐ์ˆ ์ด ์žˆ๋‹ค. ๋ชจ์…˜์บก์ณ๋Š” ํ”ผ๊ด€์ฐฐ์ž์˜ ์›€์ง์ž„์„ 3์ฐจ์› ๊ณต๊ฐ„์—์„œ ์ถ”์ ํ•˜๊ณ  ์ €์žฅํ•˜์—ฌ ๋‹ค๋ฅธ ์–ดํ”Œ๋ฆฌ์ผ€์ด์…˜์— ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ๋งŒ๋“  ๊ธฐ์ˆ ์ด๋‹ค(6).

๊ทธ๋ฆผ. 1. ์„ผ์„œ ๋ฐฐ์น˜ & IMU, ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ

Fig. 1. Sensors placement & IMU, flex pressure sensor

../../Resources/kiee/KIEE.2023.72.11.1434/fig1.png

์ด๋Ÿฌํ•œ ๋ชจ์…˜์บก์ณ ์„œ๋น„์Šค๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ์ถ”์ถœํ•˜๋Š” ๋ฐฉ์‹์— ๋”ฐ๋ผ ๊ด‘ํ•™์‹, ๊ธฐ๊ณ„์‹, ์ „์ž๊ธฐ์žฅ์‹ ๋“ฑ์œผ๋กœ ๋‹ค์–‘ํ•˜๊ฒŒ ๊ตฌํ˜„๋˜์–ด์™”๋‹ค. ๋Œ€ํ‘œ์ ์ธ ๊ด‘ํ•™์‹ ๋ชจ์…˜์บก์ณ์˜ ๊ฒฝ์šฐ ํ”ผ๊ด€์ฐฐ์ž์˜ ์‹ ์ฒด์— ๋ถ€์œ„๋งˆ๋‹ค ๋งˆ์ปค๋ฅผ ๋ถ€์ฐฉํ•œ ๋’ค ๋‹ค์ˆ˜์˜ ์นด๋ฉ”๋ผ๊ฐ€ ์ดฌ์˜ํ•œ ์˜์ƒ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์œ„์น˜๋ฅผ ์ถ”์ •ํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค(7). ๊ทธ๋ž˜์„œ ๊ด‘ํ•™์‹ ๋ชจ์…˜์บก์ณ๋Š” ๊ณ ์†์ดฌ์˜๊ณผ ๋‹ค์ˆ˜์˜ ์นด๋ฉ”๋ผ๋งŒ ํ™•๋ณด๋œ๋‹ค๋ฉด ์ •ํ™•ํ•œ ํ‘œํ˜„์ด ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ์žฅ์ ์ด ์žˆ๋‹ค(8). ํ•˜์ง€๋งŒ ๋ฐ˜๋Œ€๋กœ ์นด๋ฉ”๋ผ์˜ ๊ฐœ์ˆ˜๊ฐ€ ์ ์œผ๋ฉด ๊ด€์ฐฐ์— ์‚ฌ๊ฐ์ง€๋Œ€๊ฐ€ ๋ฐœ์ƒํ•˜์—ฌ ์ •ํ™•ํ•œ ์ถ”์ •์ด ์–ด๋ ค์›Œ์ง€๋Š” ๋“ฑ ํ™˜๊ฒฝ์ ์ธ ์š”์†Œ์— ์ œ์•ฝ๋ฐ›๋Š”๋‹ค๋Š” ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. ์ด์— ๋ฐ˜ํ•ด ๊ด€์„ฑ์„ผ์„œ ๊ธฐ๋ฐ˜์ผ ๊ฒฝ์šฐ ์ŠˆํŠธ๋ฅผ ์ฐฉ์šฉํ•˜๋ฉด ๋น„๊ต์  ๊ณต๊ฐ„์ ์ธ ์ œ์•ฝ ์—†์ด ๋ชจ์…˜์บก์ณ๋‚˜ ๋™์ž‘๋ถ„์„์„ ํ•  ์ˆ˜ ์žˆ๋‹ค(9). ๋˜ํ•œ ์‹ ์ฒด ์ „๋ฐ˜์— ๋ถ€์ฐฉ๋œ ์„ผ์„œ ๋ฐ์ดํ„ฐ๋กœ๋ถ€ํ„ฐ ์ •ํ™•ํ•œ ํŒ๋‹จ์ด ๊ฐ€๋Šฅํ•ด์ง€๋ฏ€๋กœ ์•ก์„ธ์„œ๋ฆฌํ˜• ๋””๋ฐ”์ด์Šค ๋ณด๋‹ค ์ •ํ™•ํ•œ ์šด๋™ ๋ถ„์„ ๋ฐ ๊ฐ€์ด๋“œ ์ œ๊ณต์ด ๊ฐ€๋Šฅํ•ด์ง„๋‹ค. ํ•˜์ง€๋งŒ ๋‹ค์ˆ˜์˜ IMU(Inertial Measurement Unit)๋ฅผ ์‚ฌ์šฉํ•  ๊ฒฝ์šฐ ์ œ์ž‘ ๋‹จ๊ฐ€๊ฐ€ ๋†’์•„์ง€๊ฒŒ ๋œ๋‹ค๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค.

๊ทธ๋ฆผ. 2. ์ „์ฒด ์‹œ์Šคํ…œ ๊ตฌ์กฐ๋„

Fig. 2. System architecture

../../Resources/kiee/KIEE.2023.72.11.1434/fig2.png

์ด๋Ÿฌํ•œ ์‚ฌ์šฉ ๊ณต๊ฐ„์  ํ•œ๊ณ„๋ฅผ ๊ทน๋ณตํ•จ๊ณผ ๋™์‹œ์— ์žฅ๋น„ ์ œ์ž‘ ๋น„์šฉ ์ ˆ๊ฐ์„ ๋ชฉํ‘œ๋กœ ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ํ˜„์žฌ ๊ฐœ๋ฐœ ์ค‘์ธ IMU ๋Œ€๋น„ ํ›จ์”ฌ ์ €๋ ดํ•˜๊ฒŒ ์–‘์‚ฐ ๊ฐ€๋Šฅํ•œ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ชจ์…˜์บก์ณ์™€ ๋™์ž‘๋ถ„๋ฅ˜๊ฐ€ ๊ฐ€๋Šฅํ•œ ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ์ œ์‹œํ•œ๋‹ค. ์ œ์‹œํ•œ ๋ฐฉ์‹์€ ์นด๋ฉ”๋ผ ์—†์ด ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๊ธฐ์— ๊ณต๊ฐ„์ œ์•ฝ์„ ๋ฐ›์ง€ ์•Š์œผ๋ฉฐ, ์ตœ์†Œํ•œ์˜ IMU์™€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ์กฐํ•ฉํ•˜์—ฌ ๊ตฌ์„ฑํ•˜์—ฌ ์ œ์ž‘ ๋‹จ๊ฐ€๋ฅผ ๋‚ฎ์ถœ ์ˆ˜ ์žˆ๋‹ค๋Š” ์žฅ์ ์ด ์žˆ๋‹ค. ์ œ์ž‘๋œ ์‹œ์Šคํ…œ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์ „์ด ํ•™์Šต ๊ธฐ๋ฐ˜ ์šด๋™ ๋™์ž‘ ๋ถ„๋ฅ˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๋ฐ ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•˜๋Š” ์ž์„ธ ์ถ”์ • ๋ฐ ์ •ํ™•๋„ ํ‰๊ฐ€ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌํ˜„ํ•˜์˜€๋‹ค. ๋ณธ ์—ฐ๊ตฌ์˜ ์ฃผ์š” ๊ธฐ์—ฌ ์‚ฌํ•ญ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์š”์•ฝํ•  ์ˆ˜ ์žˆ๋‹ค.

- ์‚ฌ์šฉ ๊ณต๊ฐ„์˜ ์ œ์•ฝ์„ ๋ฐ›์ง€ ์•Š๋Š” ๋น„ ๊ด‘ํ•™์‹ ๋ชจ์…˜์บก์ณ ๋ฐ ๋™์ž‘๋ถ„์„ ๊ธฐ๋ฒ•๊ตฌํ˜„

- ์žฅ๋น„ ์ œ์ž‘ ๋‹จ๊ฐ€ ์ ˆ๊ฐ์„ ์œ„ํ•ด ๋‹ค์ˆ˜์˜ IMU๋งŒ์„ ์‚ฌ์šฉํ•˜์ง€ ์•Š๊ณ  IMU์™€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜์—ฌ ์›จ์–ด๋Ÿฌ๋ธ” ์žฅ๋น„ ๊ฐœ๋ฐœ

- ์ ์€ ์ˆ˜์˜ ๋ฐ์ดํ„ฐ๋กœ๋„ ๋ถ„๋ฅ˜ ์ •ํ™•๋„๋ฅผ ๋†’์ด๊ธฐ ์œ„ํ•ด ์‚ฌ์ „ ํ•™์Šต๋œ Inception V3๋ฅผ ์ด์šฉํ•œ ์ „์ดํ•™์Šต ๊ธฐ๋ฒ• ์ ์šฉ

2. ๋ณธ ๋ก 

2.1 ํ•˜๋“œ์›จ์–ด ๊ตฌ์„ฑ

์‚ฌ์šฉ์ž์˜ ๋™์ž‘์„ ๋ถ„์„ํ•˜๊ธฐ ์œ„ํ•ด ์ œ์ž‘๋œ ํ•˜๋“œ์›จ์–ด์˜ ์„ผ์„œ ๊ตฌ์„ฑ๋„๋Š” ๊ทธ๋ฆผ 1๊ณผ ๊ฐ™๋‹ค. ์ƒ์ฒด๋ถ€ ๊ธฐ์ค€์œผ๋กœ ๋“ฑ, ์™ผํŒ” ์ƒ์™„, ์˜ค๋ฅธํŒ” ์ƒ์™„์— IMU๋ฅผ ๊ฐ 1๊ฐœ, ์–‘ํŒ” ํŒ”๊ฟˆ์น˜์— ๊ฐ๊ฐ ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ๋ฅผ 3๊ฐœ์”ฉ ๋ฐฐ์น˜ํ•˜์˜€๋‹ค. ํ•˜์ฒด๋ถ€ ๋˜ํ•œ ํ—ˆ๋ฆฌ, ์–‘์ชฝ ํ—ˆ๋ฒ…์ง€ ์•ž๋ถ€๋ถ„์— IMU๋ฅผ 1๊ฐœ์”ฉ, ์–‘์ชฝ ๋ฌด๋ฆŽ ๋ถ€๋ถ„์— ๊ด€์ ˆ ๋‹น 3๊ฐœ์˜ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ๋ฐฐ์น˜ํ•˜์˜€๋‹ค.

๋ณธ ์—ฐ๊ตฌ์— ์‚ฌ์šฉ๋œ IMU๋Š” ๋ฏธ๊ตญ InvenSense์‚ฌ์—์„œ ์ œ์ž‘ํ•œ MPU6050์ด๋‹ค. MPU6050์€ ๊ฐ ์ฑ„๋„๋‹น 16bit ADC ๋ชจ๋“ˆ์„ ๊ฐ€์ง„ MEMS(Micro Electro Mechanical Systems) ๊ธฐ๋ฐ˜ ๊ด€์„ฑ์„ผ์„œ๋กœ์จ x, y, z ์ถ•์˜ ์„ ํ˜• ๊ฐ€์†๋„, ์ž์ด๋กœ์Šค์ฝ”ํ”„ ๊ฐ’์„ ์•Œ ์ˆ˜ ์žˆ๋‹ค. ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋Š” 13cm์˜ ๊ธธ์ด๋ฅผ ๊ฐ€์กŒ์œผ๋ฉฐ ๊ตฌ๋ถ€๋Ÿฌ์ง„ ์ •๋„, ์••๋ ฅ์˜ ์ •๋„์— ๋”ฐ๋ผ ์ €ํ•ญ๊ฐ’์ด ๋ฐ”๋€Œ๋Š” ํŠน์ง•์„ ๊ฐ€์ง€๊ณ  ์žˆ๋‹ค. ์ดˆ๊ธฐ ์ €ํ•ญ์€ 10Mฮฉ์ด๋ฉฐ ๊ตฌ๋ถ€๋Ÿฌ์ง€๋Š” ์ •๋„์— ๋”ฐ๋ผ 150ฮฉ๊นŒ์ง€ ๊ฐ€๋ณ€์ด ๊ฐ€๋Šฅํ•˜๋‹ค.

IMU๋ฅผ ์‹ ์ฒด์— ์ตœ๋Œ€ํ•œ ๋ฐ€์ฐฉ์‹œํ‚ฌ ๊ฒฝ์šฐ ๊ทผ์œก์˜ ์ˆ˜์ถ•, ์ด์™„ ์ƒํƒœ์— ๋”ฐ๋ผ ๋‹ฌ๋ผ์ง€๋Š” ๊ทผ์œก์˜ ๋ชจ์–‘ ๋•Œ๋ฌธ์— ์ž์„ธ ์ถ”์ •์— ๋ฐฉํ•ด๊ฐ€ ๋  ์ˆ˜ ์žˆ๋‹ค(10). ๊ทธ๋ž˜์„œ ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ์ด๋Ÿฌํ•œ ์˜ํ–ฅ์„ ์ค„์ด๊ธฐ ์œ„ํ•ด ๊ทธ๋ฆผ 1๊ณผ ๊ฐ™์ด IMU ์„ผ์„œ๋ฅผ ํ”ผ๋ถ€์— ๋ฐ”๋กœ ๋ถ€์ฐฉํ•˜๋Š” ๋ฐฉ๋ฒ• ๋Œ€์‹  1x5.5x8์˜ ๋ถ€ํ”ผ๋ฅผ ๊ทœ๊ฒฉ์„ ๊ฐ€์ง„ ๋ธŒ๋ ˆ๋“œ๋ณด๋“œ ์œ„์— IMU ์„ผ์„œ๋ฅผ ๋ถ€์ฐฉ ํ›„ ์ฐฉ์šฉํ•˜์˜€๋‹ค.

ํ•ด๋‹น ์„ผ์„œ๋“ค์€ ์ƒ์ฒด๋ถ€์™€ ํ•˜์ฒด๋ถ€๋กœ ๋‚˜๋‰˜์–ด 2๊ฐœ์˜ ์•„๋‘์ด๋…ธ Due ๋ณด๋“œ์— ์—ฐ๊ฒฐํ–ˆ๋‹ค. MPU6050์€ I2C ํ†ต์‹ , ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋Š” ์•„๋‚ ๋กœ๊ทธ ์ž…๋ ฅ์„ ํ†ตํ•ด ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์‹ ํ•˜๋„๋ก ์ฒ˜๋ฆฌํ•˜์˜€์œผ๋ฉฐ Due ๋ณด๋“œ๋Š” PC์™€์˜ serial ํ†ต์‹ ์œผ๋กœ MATLAB ํ™˜๊ฒฝ์— ์„ผ์„œ ๋ฐ์ดํ„ฐ๋ฅผ ์ „์†กํ•˜๊ฒŒ ์ธํ„ฐํŽ˜์ด์Šค๋ฅผ ๋งž์ท„๋‹ค.

2.2 ๋™์ž‘๋ถ„๋ฅ˜ ๋ฐ ๋ชจ์…˜์บก์ณ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ์ „์ฒด ์‹œ์Šคํ…œ ๊ตฌ์กฐ

๊ทธ๋ฆผ 2๋Š” ๋ณธ ๋…ผ๋ฌธ์—์„œ ์ œ์‹œํ•˜๋Š” ์‹œ์Šคํ…œ์˜ ์ „์ฒด์ ์ธ ๊ตฌ์กฐ๋„์ด๋‹ค. ์‚ฌ์šฉ์ž์˜ ์›€์ง์ž„์„ ํ™•์ธํ•˜๋Š” ๋ณ„๋„์˜ ์นด๋ฉ”๋ผ ์—†์ด ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋งŒ์„ ์ด์šฉํ•˜์—ฌ ์‚ฌ์šฉ์ž๊ฐ€ ์–ด๋–ค ๊ทผ๋ ฅ์šด๋™์„ ํ•˜๋Š”์ง€ ๋ถ„๋ฅ˜ํ•˜๊ณ , ํ•ด๋‹น ๊ทผ๋ ฅ์šด๋™์—์„œ ์ •ํ™•ํ•œ ์ž์„ธ๋ฅผ ์ทจํ•˜๊ณ  ์žˆ๋Š”์ง€ ํŒŒ์•…ํ•˜๋Š” ์‹œ์Šคํ…œ์„ ๊ตฌ์„ฑํ•˜์˜€๋‹ค. ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ์ฐฉ์šฉํ•œ ์‚ฌ์šฉ์ž๋Š” ์›€์ง์ž„์„ ์‹œ์ž‘ํ•˜๊ธฐ ์ „์— ํŠน์ • ์ž์„ธ๋ฅผ ์ทจํ•˜์—ฌ ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜ ๊ณผ์ •์„ ์ง„ํ–‰ํ•œ๋‹ค. ์ดํ›„ ๋Ÿฐ์ง€, ์Šค์ฟผํŠธ ๋“ฑ์˜ ์šด๋™ ๋™์ž‘์„ ์‹œ์ž‘ํ•˜๋ฉด IMU์™€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์˜ ์‹ ํ˜ธ๊ฐ€ ์‹œ์Šคํ…œ์œผ๋กœ ์ „๋‹ฌ๋œ๋‹ค. ์‹œ์Šคํ…œ์€ ๋™์ž‘๋ถ„๋ฅ˜์™€ ๋ชจ์…˜์บก์ฒ˜๋กœ ๊ตฌ๋ถ„๋˜๋ฉฐ ๋™์ž‘๋ถ„๋ฅ˜๋Š” ์„ผ์„œ๋“ค์˜ ๊ฐ’์— wavelet ๋ณ€ํ™˜์„ ์ ์šฉํ•˜์—ฌ ์‹ค์‹œ๊ฐ„์œผ๋กœ ์‚ฌ์šฉ์ž๊ฐ€ ์–ด๋–ค ์šด๋™์„ ํ•˜๋Š”์ง€ ํŒŒ์•…ํ•˜๊ธฐ ์œ„ํ•ด ๊ตฌ์„ฑ๋œ๋‹ค. ๋ชจ์…˜์บก์ฒ˜๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ํ•ด๋‹น ๋™์ž‘์„ ์ •ํ™•ํ•˜๊ฒŒ ์ˆ˜ํ–‰ํ•˜๋Š”์ง€ ํŒŒ์•…ํ•˜๊ธฐ ์œ„ํ•œ ๊ฐ€์ด๋“œ๋ฅผ ์ œ์‹œํ•˜๊ธฐ ์œ„ํ•ด ๊ตฌ์„ฑ๋œ๋‹ค.

2.3 ๋ชจ์…˜์บก์ณ

๋ณธ ๋…ผ๋ฌธ์—์„œ ์ œ์‹œํ•˜๋Š” ์‹œ์Šคํ…œ์€ ์‚ฌ์šฉ์ž๊ฐ€ ๋ณ„๋„์˜ ์กฐ์ž‘ ์—†์ด ์šด๋™์„ ์‹œ์ž‘ํ•˜๋ฉด ์ฆ‰์‹œ ํ•ด๋‹น ์šด๋™์ด ๋ถ„๋ฅ˜๋˜๊ณ , ํ•ด๋‹น ์šด๋™์˜ ์ •ํ™•ํ•œ ์ž์„ธ๊ฐ€ TV ๋˜๋Š” ๊ฐœ์ธ ๋””์Šคํ”Œ๋ ˆ์ด์— ์ œ๊ณต๋œ๋‹ค. ์ด ๋•Œ ์‚ฌ์šฉ์ž๊ฐ€ ์˜ฌ๋ฐ”๋ฅธ ์ž์„ธ๋ฅผ ์œ ์ง€ํ•˜๋ฉฐ ์šด๋™ํ•˜๋Š”์ง€ ํ™•์ธํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ์ฐฉ์šฉํ•œ ์‚ฌ์šฉ์ž์˜ ์›€์ง์ž„์„ ์‹ค์‹œ๊ฐ„์œผ๋กœ ํŒŒ์•…ํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹ค.

์นด๋ฉ”๋ผ ์—†์ด ์‚ฌ์šฉ์ž์˜ ์›€์ง์ž„์„ ํŒŒ์•…ํ•˜๊ธฐ ์œ„ํ•ด IMU์™€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ ๋ชจ์…˜์บก์ณ ์‹œ์Šคํ…œ์„ ๊ตฌ์„ฑํ•œ๋‹ค. ์ด๋•Œ, ์ธ๊ฐ„์˜ ์›€์ง์ž„์„ ๊ทธ๋ฆผ 3๊ณผ ๊ฐ™์€ ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ๋กœ ๋‚˜ํƒ€๋‚ธ๋‹ค. ๊ฐ ๋ถ€์œ„์˜ ์ขŒํ‘œ๊ณ„ ํ†ต์ผ์„ ์œ„ํ•ด ์‹œ์ž‘ ์ •์ž์„ธ ๊ธฐ์ค€์œผ๋กœ ์‚ฌ์šฉ์ž ์‹ ์ฒด์˜ ์™ผ์ชฝ์„ x์ถ• ๋ฐฉํ–ฅ, ํ›„๋ฉด์„ y์ถ• ๋ฐฉํ–ฅ, ์ค‘๋ ฅ ๋ฐ˜๋Œ€ ๋ฐฉํ–ฅ์„ z์ถ• ๋ฐฉํ–ฅ์œผ๋กœ ์ „์—ญ ์ขŒํ‘œ๊ณ„๋ฅผ ์ •์˜ํ•˜์˜€์œผ๋ฉฐ IMU ์„ผ์„œ์˜ body frame๋„ ์ „์—ญ์ขŒํ‘œ์— ๋งž๊ฒŒ ์ขŒํ‘œ๋ณ€ํ™˜์„ ์ง„ํ–‰ํ•œ๋‹ค. ๋ณธ ์—ฐ๊ตฌ์˜ ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ์— ์„ค์ •๋œ ๊ด€์ ˆ์€ ๊ทธ๋ฆผ 3๊ณผ ๊ฐ™์ด ํ—ˆ๋ฆฌ, ๋“ฑ, ์–ด๊นจ, ๊ณจ๋ฐ˜, ๋ฌด๋ฆŽ, ํŒ”๊ฟˆ์น˜์ด๋‹ค. ์ƒ์ฒด๋ถ€์˜ ๋“ฑ, ์˜ค๋ฅธ์ชฝ ํŒ”, ์™ผ์ชฝ ํŒ”, ํ•˜์ฒด๋ถ€์˜ ํ—ˆ๋ฆฌ, ์˜ค๋ฅธ์ชฝ ํ—ˆ๋ฒ…์ง€, ์™ผ์ชฝ ํ—ˆ๋ฒ…์ง€ ์ˆœ์œผ๋กœ IMU ์„ผ์„œ ์ธ๋ฑ์‹ฑ์„ ํ•˜์˜€๋‹ค.

๊ทธ๋ฆผ. 3. ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ๋ง(์ •๋ฉด ๊ธฐ์ค€), IMU ์ธ๋ฑ์‹ฑ

Fig. 3. Skeleton modeling & IMU index

../../Resources/kiee/KIEE.2023.72.11.1434/fig3.png

์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค์˜ ๊ฒฝ์šฐ, ์ฐฉ์šฉ ๊ณผ์ •์—์„œ ์‚ฌ์šฉ์ž์— ๋”ฐ๋ผ ๋‹ค๋ฅธ ์ฒดํ˜•, ์˜ท์˜ ๋’คํ‹€์–ด์ง ๋“ฑ์— ์˜ํ•ด ์˜๋„์น˜ ์•Š์€ IMU์˜ ๋’คํ‹€๋ฆผ์ด ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋‹ค. ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ ๋˜ํ•œ ์˜๋ฅ˜์— ๋‚ด์žฅ๋˜์–ด ์žˆ๊ธฐ์— ์‚ฌ์šฉ์ž์˜ ์ฒดํ˜•์— ๋”ฐ๋ผ ๋ฐ€์ฐฉ ์ •๋„๊ฐ€ ๋‹ฌ๋ผ์ง€๋ฉฐ, ๊ฐ™์€ ๋™์ž‘์„ ์ˆ˜ํ–‰ํ•˜๋”๋ผ๋„ ๊ตฌ๋ถ€๋ฆผ ๊ฐ๋„ ๊ฐ’์„ ๋‚˜ํƒ€๋‚ด๋Š” ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์˜ ์ถœ๋ ฅ ์ €ํ•ญ๊ฐ’์ด ๋‹ค๋ฅผ ์ˆ˜ ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ํ˜„์ƒ์„ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด ์ดˆ๊ธฐ ๊ธฐ์šธ์–ด์ง„ IMU์˜ ๊ฐ๋„๋ฅผ ๋ณด์ƒํ•ด์ค„ ํ•„์š”๊ฐ€ ์žˆ์œผ๋ฉฐ, ์‚ฌ์šฉ์ž ์ฒดํ˜•์— ๋งž์ถฐ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์˜ ์ €ํ•ญ ์ถœ๋ ฅ ์ตœ๋Œ€, ์ตœ์†Ÿ๊ฐ’์„ ํŒŒ์•…ํ•˜๋Š” ๊ณผ์ •์ด ํ•„์š”ํ•˜๋‹ค.

๋ณธ ์—ฐ๊ตฌ์—์„œ์˜ ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜์€ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์‹œ์ž‘ 10์ดˆ๊ฐ„ ํŒ”, ๋‹ค๋ฆฌ ๊ด€์ ˆ์„ ์ตœ๋Œ€ํ•œ ๊ตฌ๋ถ€๋ฆฐ ์ž์„ธ๋ฅผ ์œ ์ง€ํ•˜๋ฉด์„œ ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ง‘ํ•œ๋‹ค. ์ด์–ด์„œ ๋‹ค์Œ 10์ดˆ๊ฐ„ ์ •์ž์„ธ๋กœ ์ •์ง€์ƒํƒœ๋ฅผ ์œ ์ง€ํ•˜๋ฉฐ ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ง‘ํ•œ๋‹ค. ์„ ํ–‰ ๋™์ž‘์—์„œ๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ํŒ”๊ฟˆ์น˜, ๋ฌด๋ฆŽ์„ ์ตœ๋Œ€ํ•œ ๊ตฌ๋ถ€๋ ธ๋‹ค ๊ฐ„์ฃผํ•˜์—ฌ ์ €์žฅ๋œ ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ์˜ ๊ฐ’์˜ ํ‰๊ท ๊ฐ’์ด ์ตœ๋Œ€ ๊ตฌ๋ถ€๋ฆผ ๊ฐ๋„(135ยฐ)์— ํ•ด๋‹นํ•œ๋‹ค๊ณ  ์ •์˜ํ•˜์˜€๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ •์ž์„ธ์—์„œ ์ˆ˜์ง‘๋œ ๊ฐ IMU ๋ฐ์ดํ„ฐ๋Š” ๋‹ค๋ฅธ ์›€์ง์ž„์ด ์—†์—ˆ๊ธฐ ๋•Œ๋ฌธ์— ์ค‘๋ ฅ๊ฐ€์†๋„์—๋งŒ ์˜ํ–ฅ์„ ๋ฐ›์•˜๋‹ค๊ณ  ๊ฐ€์ •ํ•˜์—ฌ ์‹(1), ์‹(2), ์‹(3)๊ณผ ๊ฐ™์ด ์ €์žฅ๋œ x, y, z ๊ฐ€์†๋„๋“ค์˜ ํ‰๊ท ๊ฐ’๋ผ๋ฆฌ arctan์„ ํ†ตํ•ด ๊ธฐ์šธ์–ด์ง„ ๊ฐ์„ ๊ณ„์‚ฐํ•œ๋‹ค. ๊ณ„์‚ฐ๋œ ๊ฐ๋งŒํผ ์ดˆ๊ธฐ ์ž์„ธ์— ๋ณด์ƒํ•ด์ฃผ์–ด ์ •ํ™•ํ•œ ์ •์ž์„ธ๋ฅผ ๊ตฌํ˜„ํ•˜์˜€๋‹ค. ๋˜ํ•œ ํ›„ํ–‰ ๋™์ž‘์ธ ์ •์ž์„ธ์—์„œ ์ˆ˜์ง‘๋œ ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ ๋ฐ์ดํ„ฐ๋“ค์˜ ํ‰๊ท ๊ฐ’์€ ํŒ”๊ฟˆ์น˜์™€ ๋ฌด๋ฆŽ์„ ์ตœ๋Œ€ํ•œ ํŽธ ์ƒํƒœ๋ผ๊ณ  ๊ฐ„์ฃผํ•˜์—ฌ ๊ตฌ๋ถ€๋ฆผ ๊ฐ๋„๊ฐ€ 0ยฐ์ด๋ผ๊ณ  ์ •์˜ํ•˜์˜€๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๋™์ž‘ ์‹œ ์ถœ๋ ฅ๋˜๋Š” ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์˜ ๊ฐ’์œผ๋กœ ์‚ฌ์ „์— ์ •์˜๋œ ์ตœ๋Œ€์ตœ์†Ÿ๊ฐ’ ์‚ฌ์ด ๋น„์œจ ๊ด€๊ณ„๋ฅผ ์ด์šฉํ•ด 0ยฐ~135ยฐ์˜ ํŒ”๊ฟˆ์น˜, ๋ฌด๋ฆŽ ๊ฐ๋„๋ฅผ ์ถ”์ •ํ•˜์˜€๋‹ค.

(1)
$\rho=\arctan \left(\frac{A_x}{\sqrt{A_y^2+A_z^2}}\right)$

(2)
$\phi=\arctan \left(\frac{A_y}{\sqrt{A_x^2+A_z^2}}\right)$

(3)
$\Theta=\arctan \left(\frac{\sqrt{A_x^2+A_y^2}}{A_z}\right)$

์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜ ์ดํ›„ ์‚ฌ์šฉ์ž๊ฐ€ ์›€์ง์ด๋Š” ๊ณผ์ •์—์„œ ๋ฐ”๋€Œ๋Š” ๊ด€์ ˆ์˜ ๊ฐ๋„๋Š” ํ•ด๋‹น ๋ถ€์œ„์— ๋ถ€์ฐฉ๋œ IMU์˜ ์ž์„ธ ๊ธฐ์šธ๊ธฐ๋ฅผ ์ถ”์ •ํ•จ์œผ๋กœ์จ ์•Œ ์ˆ˜ ์žˆ๋‹ค. IMU์—์„œ ์ถœ๋ ฅ๋˜๋Š” ์„ ํ˜• ๊ฐ€์†๋„์™€ ์ž์ด๋กœ์Šค์ฝ”ํ”„ ๊ฐ’์˜ ์œตํ•ฉํ•ด์„œ ์ž์„ธ๊ฐ์„ ์ถ”์ •ํ•  ์ˆ˜ ์žˆ๋Š”๋ฐ ๋Œ€ํ‘œ์ ์ธ ๋ฐฉ๋ฒ•์ด ์ƒ๋ณด ํ•„ํ„ฐ์ด๋‹ค(11). ์ƒ๋ณดํ•„ํ„ฐ๋Š” ๊ฐ€์†๋„ ๊ฐ’๋งŒ์œผ๋กœ ๊ฐ๋„๋ฅผ ๊ตฌํ–ˆ์„ ๋•Œ ๋…ธ์ด์ฆˆ์™€ ์ง„๋™์— ์ทจ์•ฝํ•˜๋‹ค๋Š” ๋‹จ์ ๊ณผ ์ž์ด๋กœ์Šค์ฝ”ํ”„ ๊ฐ’๋งŒ์œผ๋กœ ๊ฐ๋„๋ฅผ ๊ตฌํ–ˆ์„ ๋•Œ ๋“œ๋ฆฌํ”„ํŠธ๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค๋Š” ๋‹จ์ ์„ ์ƒํ˜ธ๋ณด์™„ํ•œ ํ•„ํ„ฐ๋ง ๊ธฐ๋ฒ•์œผ๋กœ์จ ์‹(4)์™€ ๊ฐ™์ด ์ •์˜๋œ๋‹ค.

(4)
$\hat{\theta}=\alpha \theta_{\text {gy ro }}+(1-\alpha) \theta_{\text {acc }}$

๊ณ ์ฃผํŒŒ์—์„œ ์‘๋‹ต ํŠน์„ฑ์ด ์ข‹์€ ์ž์ด๋กœ์Šค์ฝ”ํ”„ ๊ฐ’์€ ๊ณ ์ฃผํŒŒ ํ•„ํ„ฐ๋ฅผ ํ†ต๊ณผ์‹œํ‚ค๊ณ  ์ €์ฃผํŒŒ์—์„œ ์‘๋‹ต ํŠน์„ฑ์ด ์ข‹์€ ๊ฐ€์†๋„ ๊ฐ’์€ ์ €์ฃผํŒŒ ํ•„ํ„ฐ๋ฅผ ์ ์šฉ์‹œํ‚จ ๋’ค ๊ฐ๊ฐ ๊ฐ€์ค‘์น˜ ฮฑ์™€ (1-ฮฑ)๋ฅผ ๊ณฑํ•˜๊ณ  ๋‘ ๊ฐ’์„ ๋”ํ•œ๋‹ค. ์ด๋•Œ ฮฑ ๊ฐ’์— ๋”ฐ๋ผ ํ•„ํ„ฐ์˜ ํŠน์„ฑ์ด ๋ฐ”๋€Œ๋Š”๋ฐ ฮฑ ๊ฐ’์ด ์ปค์งˆ์ˆ˜๋ก ์ž์ด๋กœ์Šค์ฝ”ํ”„์˜ ๋น„์ค‘์ด ์ปค์ง€๋ฏ€๋กœ ์•ˆ์ •์ ์ธ ์ถœ๋ ฅ์ด ๋‚˜์˜ค์ง€๋งŒ ๋ฐ˜์‘์ด ๋Š๋ฆฌ๋‹ค๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค. ์ด์— ๋ฐ˜ํ•ด ฮฑ๊ฐ’ ์ž‘์•„ ์ง€๋ฉด ๊ฐ€์†๋„์˜ ๋น„์ค‘์ด ์ปค์ง€๋ฏ€๋กœ ๋…ธ์ด์ง€ํ•œ ์ถœ๋ ฅ์„ ๋ณด์ด์ง€๋งŒ ๋น ๋ฅด๊ฒŒ ๋ฐ˜์‘ํ•œ๋‹ค๋Š” ์žฅ์ ์ด ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ํŠน์„ฑ์„ ๊ณ ๋ คํ•˜์—ฌ ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๋น„๊ต์  ์ด๋™๋ฐ˜๊ฒฝ์ด ์ž‘์ง€๋งŒ ์ƒยทํ•˜์ฒด์˜ ์ค‘์‹ฌ์ด ๋˜๋Š” 2๋ฒˆ, 5๋ฒˆ IMU์˜ ์ƒ๋ณดํ•„ํ„ฐ์—๋Š” ฮฑ ๊ฐ’์„ 0.99๋กœ, ํŒ”๋‹ค๋ฆฌ ๋ถ€๋ถ„์„ ๋‹ด๋‹นํ•˜์—ฌ ๋งŽ์€ ์›€์ง์ž„์ด ์˜ˆ์ƒ๋˜๋Š” ๋‚˜๋จธ์ง€ 4๊ฐœ์˜ IMU์˜ ์ƒ๋ณดํ•„ํ„ฐ์—๋Š” ฮฑ ๊ฐ’์„ 0.65๋กœ ์„ค์ •ํ•˜์˜€๋‹ค.

๋ฐฉํ–ฅ ์ฝ”์‚ฌ์ธ ํ–‰๋ ฌ(DCM)์€ ์„œ๋กœ ๋‹ค๋ฅธ ๋‘ ์ขŒํ‘œ๊ณ„์˜ ๋‹จ์œ„ ๋ฒกํ„ฐ ๊ฐ„์— ๋‚ด์ ๋“ค์ด ์„ฑ๋ถ„์œผ๋กœ ์ด๋ฃจ์–ด์ง„ ํ–‰๋ ฌ์ด๋ฉฐ ๋ฒกํ„ฐ์˜ ์ขŒํ‘œ๋ณ€ํ™˜์˜ ๋ฐฉ์‹์œผ๋กœ ์‚ฌ์šฉ๋œ๋‹ค. ์ƒ๋ณดํ•„ํ„ฐ์—์„œ ์ถœ๋ ฅ๋˜๋Š” ๊ฐ๋„๋Š” ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ์˜ ์‹œ์ž‘ ์ž์„ธ๋ฅผ ๊ธฐ์ค€์œผ๋กœ ํ•œ ์˜ค์ผ๋Ÿฌ ๊ฐ๋„์ด๊ธฐ ๋•Œ๋ฌธ์— ์ด ๊ฐ๋„๋ฅผ Euler - Direct cosine matrix ๋ณ€ํ™˜์„ ๊ฑฐ์ณ ๋ฐฉํ–ฅ ์ฝ”์‚ฌ์ธ ํ–‰๋ ฌ์„ ๊ตฌํ•˜๋ฉด ๊ทธ๋ฆผ 3์— ์ •์˜ํ•œ ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ์— forward kinematics๋ฅผ ์‰ฝ๊ฒŒ ์ ์šฉํ•  ์ˆ˜ ์žˆ๋‹ค.

(5)
\begin{align*} R_{xyz}= R_{x}(\phi)R_{y}(\theta)R_{z}(\psi)\\ =\begin{bmatrix}c_{\theta}c_{\psi}&c_{\theta}s_{\psi}&-s_{\theta}\\s_{\phi}s_{\theta}c_{\psi}-c_{\theta}s_{\psi}&s_{\phi}s_{\theta}s_{\psi}-c_{\phi}c_{\psi}&c_{\theta}s_{\phi}\\c_{\phi}s_{\theta}c_{\psi}-c_{\theta}s_{\psi}&c_{\phi}s_{\theta}s_{\psi}-s_{\phi}c_{\psi}&c_{\theta}c_{\phi}\end{bmatrix} \end{align*}

๊ด€์ ˆ์ด ๋ฒŒ์–ด์ง„ ๊ฐ๋„์™€ ๊ฐ ๋ถ€์œ„์˜ ๊ธธ์ด๋ฅผ ์•Œ ๋•Œ ๊ทธ ๋ ์ง€์ ์„ ์•Œ ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ์—์„œ ์ •๊ธฐ๊ตฌํ•™์˜ ๊ฐœ๋…๊ณผ ๊ฐ™๋‹ค. ์ƒ์ฒด๋ฅผ ๊ธฐ์ค€์œผ๋กœ ์ค‘์‹ฌ์ด ๋˜๋Š” ๋“ฑ์˜ ์ž์„ธ๋ฅผ ๋จผ์ € ๊ตฌํ•˜๋ฉด ๋ ์ง€์ ์ด ํ•˜์œ„ ๊ด€์ ˆ์ธ ์–ด๊นจ์˜ ์‹œ์ž‘์ ์ด๋ผ๊ณ  ํ•  ์ˆ˜ ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ตฌ์กฐ๋กœ ์†๋ชฉ ๋ฐœ๋ชฉ๊นŒ์ง€ ์—ฐ๊ฒฐํ•ด ๋‚˜๊ฐ€๋ฉด ์‚ฌ์šฉ์ž์˜ ์›€์ง์ž„์„ ์ถ”์ข…ํ•  ์ˆ˜ ์žˆ๋‹ค.

๋ณธ ๋…ผ๋ฌธ์—์„œ ์ œ์•ˆํ•œ ์‹œ์Šคํ…œ์€ ์ƒ์ฒด 3๊ฐœ, ํ•˜์ฒด 3๊ฐœ ์ด 6๊ฐœ์˜ IMU๋ฅผ ์ด์šฉํ•˜์—ฌ ๊ตฌ์„ฑํ•˜์˜€๊ธฐ์—, ํŒ”๊ฟˆ์น˜์™€ ๋ฌด๋ฆŽ ๊ด€์ ˆ ์ดํ›„์˜ ์›€์ง์ž„์€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋ฅผ ์ด์šฉํ•˜์—ฌ forward kinematics๋ฅผ ์ ์šฉํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹ค. ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ์—์„œ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๊ฐ€ ๊ด€์—ฌํ•˜๋Š” ํŒ”๊ฟˆ์น˜์™€ ๋ฌด๋ฆŽ์˜ ๊ฐ๋„๋Š” ์˜ค์ผ๋Ÿฌ ์ขŒํ‘œ ๊ธฐ์ค€์œผ๋กœ ์ƒ์œ„ ๊ด€์ ˆ์ธ ํŒ”๊ฟˆ์น˜์™€ ํ—ˆ๋ฒ…์ง€์˜ ๋ถ€์ฐฉ๋œ IMU ์„ผ์„œ์˜ ๊ด€์„ฑ ์ขŒํ‘œ๊ณ„ ๊ธฐ์ค€ y์ถ•๊ณผ z์ถ•์€ ๋™์ผํ•˜๊ณ  x์ถ•๋งŒ ๋‹ค๋ฅด๋‹ค. ์ด๋ฅผ ์ด์šฉํ•˜์—ฌ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์˜ ์ €ํ•ญ ์ถœ๋ ฅ๊ฐ’์„ ์ด์šฉํ•˜์—ฌ ์ดˆ๊นƒ๊ฐ’ ๋Œ€๋น„ ๋ณ€ํ™”๋œ ์–‘๋งŒํผ x์ถ•์˜ ํšŒ์ „์„ ๊ณ ๋ คํ•ด ์ค„ ์ˆ˜ ์žˆ๋‹ค.

2.4 ๋™์ž‘๋ถ„๋ฅ˜

๋ณธ ๋…ผ๋ฌธ์—์„œ ์ œ์‹œํ•˜๋Š” ๋™์ž‘๋ถ„๋ฅ˜๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ๋ณ„๋„์˜ ์กฐ์ž‘ ์—†์ด ์šด๋™์„ ์‹œ์ž‘ํ•˜๋ฉด ์ฆ‰์‹œ ํ•ด๋‹น ์šด๋™์ด ๋ถ„๋ฅ˜ํ•˜๊ธฐ ์œ„ํ•ด ๊ตฌ์„ฑ๋œ๋‹ค. ๋‹ค์–‘ํ•œ ๊ทผ๋ ฅ์šด๋™์„ ์ˆœ์ฐจ์ ์œผ๋กœ ํ•  ๋•Œ ๊ธฐ์กด ์šด๋™ ๊ฐ€์ด๋“œ ์‹œ์Šคํ…œ์˜ ๊ฒฝ์šฐ ์šด๋™ ์ข…๋ฅ˜๋ฅผ ๋ฐ”๊ฟ€ ๋•Œ๋งˆ๋‹ค ์‚ฌ์šฉ์ž๊ฐ€ ์–ด๋–ค ๊ทผ๋ ฅ์šด๋™์„ ํ• ์ง€ ์„ ํƒํ•˜๋Š” ๊ณผ์ •์ด ์žˆ์–ด์•ผ ํ•œ๋‹ค. ์ด๋Ÿฌํ•œ ๊ณผ์ •์„ ์—†์• ๊ณ  ์‚ฌ์šฉ์ž๊ฐ€ ๋ฉˆ์ถค ์—†์ด ๊ทผ๋ ฅ์šด๋™์„ ์ˆœ์ฐจ์ ์œผ๋กœ ํ•  ์ˆ˜ ์žˆ๋„๋ก ์‚ฌ์šฉ์ž๊ฐ€ ํ˜„์žฌ ์ˆ˜ํ–‰ํ•˜๊ณ  ์žˆ๋Š” ์šด๋™์„ ์‹ค์‹œ๊ฐ„์œผ๋กœ ๋ถ„๋ฅ˜ํ•˜๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌ์„ฑํ•œ๋‹ค. ๋ถ„๋ฅ˜ํ•  ์šด๋™ ๋™์ž‘์œผ๋กœ๋Š” ๊ฐ ๋ถ€์œ„๋ณ„ ์›€์ง์ž„์„ ๋‹ค์–‘ํ•˜๊ฒŒ ๋ฐ˜์˜ํ•  ์ˆ˜ ์žˆ๋„๋ก ์Šค์ฟผํŠธ, ๋ฒ„ํ”ผ, ๋Ÿฐ์ง€, ํŠธ์œ„์ŠคํŠธ, ๋ค๋ฒจ ํ”„๋ ˆ์Šค๋ฅผ ์„ ์ •ํ•˜์˜€๋‹ค.

์˜๋ฅ˜ ์ผ์ฒดํ˜• ์ฐฉ์šฉํ˜• ๋””๋ฐ”์ด์Šค๋ฅผ ์ฐฉ์šฉํ•œ ์‚ฌ์šฉ์ž๊ฐ€ ์šด๋™์„ ์‹œ์ž‘ํ•˜๋ฉด 6๊ฐœ์˜ IMU์—์„œ๋Š” ๊ฐ€์†๋„์™€ ์ž์ด๋กœ์Šค์ฝ”ํ”„ ์‹ ํ˜ธ๊ฐ€ ์ถœ๋ ฅ๋œ๋‹ค. IMU ์„ผ์„œ๋Š” ์ถœ๋ ฅ์‹ ํ˜ธ์˜ ๋…ธ์ด์ฆˆ๊ฐ€ ํฌ๊ณ  ์ œํ’ˆ๋งˆ๋‹ค ์ƒ์ดํ•œ ๋ฐ”์ด์–ด์Šค์™€ scale factor๋ฅผ ๊ฐ€์ง„๋‹ค๋Š” ํŠน์ง•์ด ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ์ ์„ ๊ณ ๋ คํ•˜์—ฌ ์ฃผํŒŒ์ˆ˜ ์˜์—ญ์—์„œ์˜ ์‹ ํ˜ธ ๋ถ„์„์„ ํ†ตํ•ด ์‹ ํ˜ธ์˜ ๋งฅ๋ฝ์ ์ธ ์ •๋ณด๋ฅผ ์ถ”๊ฐ€๋กœ ํŒŒ์•…ํ•  ํ•„์š”๊ฐ€ ์žˆ์—ˆ๊ณ  ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์„ ๊ทธ ๋ฐฉ๋ฒ•์œผ๋กœ ์‚ฌ์šฉํ•˜์˜€๋‹ค. ์—ฐ์† ์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์€ ์ž„์˜์˜ ์‹œ๊ฐ„์˜์—ญ ์‹ ํ˜ธ๋ฅผ ์›จ์ด๋ธ”๋ฆฟ ํ•จ์ˆ˜๋ฅผ ๊ธฐ์ €๋กœ ์‚ฌ์šฉํ•˜์—ฌ ์ฃผํŒŒ์ˆ˜ ์„ฑ๋ถ„์„ ๋ถ„์„ํ•˜๋Š” ๋ฐฉ๋ฒ•์ด๋ฉฐ ์‹(6)๊ณผ ๊ฐ™์ด ํ‘œํ˜„ํ•  ์ˆ˜ ์žˆ๋‹ค(12).

(6)
$T(a,\:b)=\dfrac{1}{\sqrt{a}}\int_{-\infty}^{\infty}x(t)\psi *\dfrac{t-b}{a}dt$

$x(t)$๋Š” ์‹œ๊ฐ„์˜์—ญ์˜ ์‹ ํ˜ธ์ด๋ฉฐ a์™€ b๋Š” ์ž…๋ ฅ ์‹ ํ˜ธ์˜ scaling ๋ฐ shifting์„ ์œ„ํ•œ ํŒŒ๋ผ๋ฏธํ„ฐ๋“ค์ด๊ณ , ฯˆ๊ฐ€ ์›จ์ด๋ธ”๋ฆฟ ํ•จ์ˆ˜์ด๋‹ค. ์›จ์ด๋ธ”๋ฆฟ ํ•จ์ˆ˜๋Š” ์ •ํ•ด์ง„ ์‹œ๊ฐ„ ๋™์•ˆ ์ฆ๊ฐ€์™€ ๋ฐ˜๋ณต์„ ๋ฐ˜๋ณตํ•˜์—ฌ ํ‰๊ท ์ด 0์ธ ์‹ ํ˜ธ์ด๋ฉฐ ๋Œ€ํ‘œ์ ์œผ๋กœ Mexican Hat, Morlet, Biorthoganal ๋“ฑ์ด ์žˆ๋‹ค. ์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์€ ํ‘ธ๋ฆฌ์— ๋ณ€ํ™˜๊ณผ ๋น„์Šทํ•œ ์—ญํ• ์„ ํ•˜์ง€๋งŒ ํ‘ธ๋ฆฌ์— ๋ณ€ํ™˜์€ ์‹ ํ˜ธ๊ฐ€ ์‹œ๊ฐ„์— ๋Œ€ํ•ด์„œ ๋ณ€ํ•˜์ง€ ์•Š๋Š”๋‹ค๋Š” ๊ฐ€์ •ํ•˜์— ๋‹จ์ผ ์ฃผํŒŒ์ˆ˜์— ๋Œ€ํ•œ ์ •๋ณด๋งŒ์œผ๋กœ ์‹ ํ˜ธ๋ฅผ ๋ถ„์„ํ•˜๋Š” ๋ฐ˜๋ฉด ์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์€ ์ž‘์€ ์ฃผํŒŒ์ˆ˜์—์„  ๋„“์€ ์œˆ๋„์šฐ๋กœ ํฐ ์ฃผํŒŒ์ˆ˜์—์„  ์ข์€ ์œˆ๋„์šฐ๋กœ ์Šค์ผ€์ผ ๋ณ€ํ™”์‹œํ‚ค๋ฉฐ ์—ฐ์‚ฐํ•œ๋‹ค. ๊ทธ๋Ÿฌ๋ฏ€๋กœ ์‹ ํ˜ธ๊ฐ€ ์‹œ๊ฐ„์— ๋”ฐ๋ผ ์ฃผํŒŒ์ˆ˜ ์„ฑ๋ถ„์ด ๋ณ€ํ™”ํ•˜๋”๋ผ๋„ ์ •ํ™•ํ•œ ๋ถ„์„์„ ํ•  ์ˆ˜ ์žˆ๋‹ค(13).

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๊ฐ๊ฐ์˜ IMU์—์„œ ์ถœ๋ ฅ๋˜๋Š” ์„ ํ˜• ๊ฐ€์†๋„ ๋ฐ์ดํ„ฐ์— $a_{x},\: a_{y},\: a_{z}$๋Œ€ํ•ด ์—ฐ์† ์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์‹œ์ผฐ๊ณ  ์ƒ์„ฑ๋œ ์ด๋ฏธ์ง€๋ฅผ R, G, B ์ฑ„๋„์— ์ˆœ์„œ๋Œ€๋กœ ๋ฐฐ์ •ํ•˜์—ฌ 3์ฐจ์› ์ด๋ฏธ์ง€๋ฅผ ๋งŒ๋“ค์—ˆ๋‹ค. ์ž์ด๋กœ์Šค์ฝ”ํ”„ ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•ด์„œ๋„ ์ด ๊ณผ์ •์„ ๋™์ผํ•˜๊ฒŒ ์ ์šฉํ•˜์˜€๋‹ค. ์ƒ์„ฑ๋œ 12๊ฐœ์˜ ์ด๋ฏธ์ง€๋ฅผ ํ•ฉ์ณ ํ•˜๋‚˜์˜ ์ƒˆ๋กœ์šด ์ž…๋ ฅ ์ด๋ฏธ์ง€๋ฅผ ์ •์˜ํ•ด์ฃผ์—ˆ์œผ๋ฉฐ ์ž…๋ ฅ์˜ ๋ฐฐ์—ด๊ณผ ๋™์ž‘์— ๋”ฐ๋ฅธ ์ด๋ฏธ์ง€ ๊ฒฐ๊ณผ๋Š” ๊ทธ๋ฆผ 4์™€ ๊ฐ™๋‹ค.

๊ทธ๋ฆผ. 4. ์Šค์ผˆ๋ ˆํ†ค ๋ชจ๋ธ๋ง(์ •๋ฉด ๊ธฐ์ค€), IMU ์ธ๋ฑ์‹ฑ

Fig. 4. Skeleton modeling & IMU index

../../Resources/kiee/KIEE.2023.72.11.1434/fig4.png

๊ทธ๋ฆผ 4๋ฅผ ํ†ตํ•ด ๋น„๊ต์  ํŒ” ๋™์ž‘๋งŒ์œผ๋กœ ์ˆ˜ํ–‰๋˜๋Š” ๋ค๋ฒจ ๋™์ž‘์˜ ๊ฒฝ์šฐ ์ด๋ฏธ์ง€ ์ƒ๋‹จ๋ถ€์— ํŠน์ง•์ด ๋‚˜ํƒ€๋‚˜๊ณ  ์ „์‹ ์šด๋™์ธ ๋ฒ„ํ”ผ์˜ ๊ฒฝ์šฐ ์ด๋ฏธ์ง€ ์ „๋ฐ˜์ ์œผ๋กœ ์‹ ํ˜ธ๊ฐ€ ํ™œ์„ฑํ™”๋œ ๊ฒƒ์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค. ๋˜ํ•œ ์ƒ์ฒด๋ฅผ ์œ ์ง€ํ•˜๊ณ  ํ•˜์ฒด๋ฅผ ๊ตฌ๋ถ€๋ฆฌ๋Š” ๋™์ž‘์ธ ์Šค์ฟผํŠธ์™€ ๋Ÿฐ์ง€์˜ ๊ฒฝ์šฐ ์ƒ์ฒด์˜ IMU๋Š” ์œ ์‚ฌํ•œ ํŒจํ„ด์œผ๋กœ ์ถœ๋ ฅ์ด ๋‚˜์˜ค์ง€๋งŒ, ํ•˜์ฒด์˜ IMU๊ฐ€ ๋‹ค๋ฅธ ํŒจํ„ด์„ ๋ณด์—ฌ์ฃผ๊ณ  ์žˆ์Œ์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค.

์›จ์ด๋ธ”๋ฆฟ ๋ณ€ํ™˜์„ ํ†ตํ•ด ์ด๋ฏธ์ง€๋กœ ์žฌ์ƒ์„ฑ ๋œ ์„ผ์„œ์˜ ์‹ ํ˜ธ๋ฅผ CNN (convolution neural network)๋ฅผ ํ†ตํ•ด ๋ถ„๋ฅ˜ํ•˜์˜€๋‹ค. ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์€ ๋ฐฉ๋Œ€ํ•œ ํŒŒ๋ผ๋ฏธํ„ฐ๋“ค์„ ๊ณ ๋ คํ•˜์—ฌ ์‚ฌ์ „์— ๋งŽ์€ ์–‘์˜ ๋ฐ์ดํ„ฐ๋ฅผ ํ™•๋ณดํ•ด์•ผ ํ•˜๋Š”๋ฐ ๋Œ€๋ถ€๋ถ„์˜ ๊ฒฝ์šฐ ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ง‘ํ•˜๋Š” ๊ณผ์ •์—์„œ ํฐ ๋น„์šฉ๊ณผ ์ž์›์ด ์š”๊ตฌ๋œ๋‹ค. ์ด๋Ÿฌํ•œ ๋ฌธ์ œ์˜ ํ•ด๊ฒฐ์ฑ…์œผ๋กœ ํŠน์ • ํ™˜๊ฒฝ, ๋ชฉ์  ์•„๋ž˜์„œ ์ด๋ฏธ ํ•™์Šต์ด ์ด๋ฃจ์–ด์ง„ ๋ชจ๋ธ์˜ ์ผ๋ถ€ ํ˜น์€ ์ „์ฒด๋ฅผ ๊ฐ€์ ธ์™€ ๋‹ค๋ฅธ ๋ถ„์•ผ์— ์‘์šฉํ•˜์—ฌ ํšจ๊ณผ๋ฅผ ๋ณผ ์ˆ˜ ์žˆ๋Š” ์ „์ด ํ•™์Šต์ด ์‚ฌ์šฉ๋œ๋‹ค. ๋ณธ ์—ฐ๊ตฌ์—์„œ๋„ ๋‹ค์–‘ํ•œ ์‚ฌ์šฉ์ž์˜ ๋™์ž‘์œผ๋กœ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ์…‹์„ ํ™•๋ณดํ•˜์ง€ ๋ชปํ•œ ์กฐ๊ฑด์—์„œ ์ตœ๋Œ€ํ•œ ๋†’์€ ์„ฑ๋Šฅ์„ ๋งŒ๋“ค์–ด ๋‚ด๊ธฐ ์œ„ํ•ด ์ „์ด ํ•™์Šต์„ ์‚ฌ์šฉํ•˜์˜€๋‹ค(14).

์‚ฌ์šฉ๋œ ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์€ Inception v3์ด๋ฉฐ Inception v3๋Š” Google์—์„œ ์ œ์ž‘ํ•œ ๋„คํŠธ์›Œํฌ๋กœ์จ GoogLeNet์œผ๋กœ ์•Œ๋ ค์ง„ Inception v1์„ ๊ฐœ์„ ํ•œ ๋ชจ๋ธ์ด๋‹ค. ๋น„๋Œ€์นญ ๊ตฌ์กฐ์™€ ํ•ฉ์„ฑ๊ณฑ ๋ถ„ํ•ด๋กœ ์—ฐ์‚ฐ๋Ÿ‰๊ณผ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ์ค„์ด๋ฉด์„œ ์ •ํ™•์„ฑ์„ ๋†’์˜€๋‹ค(15).

MATLAB ํ™˜๊ฒฝ์—์„œ Deep Learning Toolbox์„ ์ด์šฉํ•˜์—ฌ ๊ฐœ๋ฐœ์„ ์ง„ํ–‰ํ•˜์˜€์œผ๋ฉฐ ํ›ˆ๋ จ๋ฐ์ดํ„ฐ๋Š” ๋™์ž‘ ํด๋ž˜์Šค๋ณ„๋กœ ์•ฝ 100๊ฐœ ์ •๋„ ํ™•๋ณดํ•˜์˜€๊ณ  train/validation ๋น„์œจ์€ 8.5:1.5๋กœ ์ง€์ •ํ•˜์˜€๋‹ค. ์ดˆ๊ธฐ learning rate๋Š” 0.005๋กœ ์„ค์ •ํ•˜์˜€์œผ๋ฉฐ Minibatch size๋Š” 20์œผ๋กœ ํ•˜๊ณ  ์ ์€ ๋ฐ์ดํ„ฐ์…‹์˜ ๊ทœ๋ชจ๋ฅผ ๊ณ ๋ คํ•˜์—ฌ epoch๋Š” 30์œผ๋กœ ํ•˜์˜€๋‹ค. ํ•™์Šต ๊ฒฐ๊ณผ 92.22%์˜ ๊ฒ€์ฆ ์ •ํ™•๋„๋ฅผ ๋‚˜ํƒ€๋‚ด์—ˆ๋‹ค.

CNN ๋ชจ๋ธํ•™์Šต์„ ์œ„ํ•ด ์ž„์˜์˜ ์‚ฌ์šฉ์ž๊ฐ€ ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ์ฐฉ์šฉํ•˜๊ณ  ํ›ˆ๋ จ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ง‘ํ•˜๋Š” ๊ณผ์ •์„ ์ง„ํ–‰ํ–ˆ๋‹ค. ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘ ์‹œ ์ž„์˜์˜ ์‚ฌ์šฉ์ž๋Š” 2์ดˆ์˜ ์‹œ๊ฐ„ ๊ฐ„๊ฒฉ์„ ๋‘๊ณ  ๊ฐ ๋™์ž‘ ๋ณ„๋กœ 1ํšŒ์˜ ๋™์ž‘์„ ์ˆ˜ํ–‰ํ•˜๊ฒŒ ๋œ๋‹ค. ํ•˜์ง€๋งŒ ๋ช…ํ™•ํ•˜๊ฒŒ ๊ตฌ๋ถ„๋œ ํ›ˆ๋ จ๋ฐ์ดํ„ฐ ๋Œ€๋น„ ์‹ค์ œ ์‹œ์Šคํ…œ์„ ์‚ฌ์šฉํ•  ๋•Œ๋Š” ์—ฐ์†์ ์ธ ๋™์ž‘ ๋ณ€ํ™”๊ฐ€ ๋ฐœ์ƒํ•˜๊ฒŒ ๋˜๋ฉฐ, ์ด๋กœ ์ธํ•ด ์„ผ์„œ๋กœ๋ถ€ํ„ฐ ์—ฐ์†๋œ ์ž…๋ ฅ์ด ๋“ค์–ด์™”์„ ๋•Œ์—๋Š” ๋ถ„๋ฅ˜ ์ •ํ™•๋„๊ฐ€ ๋‚ฎ์•„์ง€๊ธฐ๋„ ํ•œ๋‹ค. ์ด๋ฅผ ๋ฐฉ์ง€ํ•˜๊ธฐ ์œ„ํ•ด ๋™์ž‘ ๋ถ„๋ฅ˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ตœ์ข… ์ถœ๋ ฅ๋‹จ์— ์Šฌ๋ผ์ด๋”ฉ ์œˆ๋„์šฐ๋ฅผ ๊ตฌ์„ฑํ–ˆ๋‹ค. 5๋งŒํผ์˜ ์‚ฌ์ด์ฆˆ๋ฅผ ๊ฐ€์ง„ ์œˆ๋„์šฐ๊ฐ€ ๊ฐ™์ด ์ด๋™ํ•˜๋ฉฐ ์œˆ๋„์šฐ ๋‚ด์—์„œ ์ตœ๋นˆ๊ฐ’์„ ํ•ด๋‹น ๋™์ž‘์œผ๋กœ ์ถœ๋ ฅํ•˜๋„๋ก ๊ทธ๋ฆผ 5์™€ ๊ฐ™์ด ์„ค๊ณ„ํ•˜์˜€๋‹ค.

๊ทธ๋ฆผ. 5. ์Šฌ๋ผ์ด๋”ฉ ์œˆ๋„์šฐ๋ฅผ ํ†ตํ•œ ๋ถ„๋ฅ˜

Fig. 5. Classification via sliding window

../../Resources/kiee/KIEE.2023.72.11.1434/fig5.png

๋˜ํ•œ ์šด๋™ ๋™์ž‘์˜ ์ •ํ™•์„ฑ ํŒ๋‹จ ๊ธฐ์ค€์€ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๊ฐ€ ๋‚˜ํƒ€๋‚ด๋Š” ํŒ”๊ฟˆ์น˜์™€ ๋ฌด๋ฆŽ์˜ ๊ตฌ๋ถ€๋ฆผ ๊ฐ๋„๋ฅผ ๊ณ ๋ คํ•˜์—ฌ ์„ค๊ณ„ํ•˜์˜€๋‹ค. ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋ฅผ ์ฐฉ์šฉํ•œ ์‚ฌ์šฉ์ž๊ฐ€ ๊ธฐ์ค€ ๊ฐ์— ๋„๋‹ฌํ•˜์ง€ ๋ชปํ•œ ์ฑ„๋กœ ํ•ด๋‹น ๋™์ž‘์„ ์ˆ˜ํ–‰ํ•˜๋ฉด ํšŸ์ˆ˜์— ๋ฐ˜์˜ํ•˜์ง€ ์•Š๋„๋ก ํ•˜์˜€๋‹ค. ๊ฐ ๋™์ž‘ ๋ณ„ ๊ธฐ์ค€ ๊ฐ๋„๋Š” ํ‘œ 1๊ณผ ๊ฐ™๋‹ค.

ํ‘œ 1. ๊ฐ€์ด๋“œ๋ฅผ ์œ„ํ•œ ๋™์ž‘๋ณ„ ๊ธฐ์ค€ ๊ฐ๋„

Table 1. Reference angle by motion for guide

์šด๋™ ๋™์ž‘

ํšŸ์ˆ˜ ์ธ์ • ๊ธฐ์ค€ ๊ฐ๋„[deg]

๋ค๋ฒจ

45 (ํŒ”๊ฟˆ์น˜)

์Šค์ฟผํŠธ

70 (๋ฌด๋ฆŽ)

๋Ÿฐ์ง€

55 (๋ฌด๋ฆŽ)

3. ์‹คํ—˜ ๋ฐ ๊ฒฐ๊ณผ

๋ณธ ๋…ผ๋ฌธ์—์„œ ๊ตฌ์„ฑํ•œ ์˜๋ฅ˜ ์ผ์ฒดํ˜• ์›จ์–ด๋Ÿฌ๋ธ” ๋””๋ฐ”์ด์Šค๋Š” ์ƒ์ฒด๋ถ€ ๊ธฐ์ค€์œผ๋กœ ๋“ฑ, ์™ผํŒ” ์ƒ์™„, ์˜ค๋ฅธํŒ” ์ƒ์™„์— MPU6050 IMU๋ฅผ ๊ฐ 1๊ฐœ, ์–‘ ํŒ” ํŒ”๊ฟˆ์น˜์— ๊ฐ๊ฐ ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ๋ฅผ 3๊ฐœ์”ฉ ๋ฐฐ์น˜๋˜์–ด ์žˆ๋‹ค. ํ•˜์ฒด๋ถ€ ๋˜ํ•œ ํ—ˆ๋ฆฌ, ์–‘์ชฝ ํ—ˆ๋ฒ…์ง€ ์•ž๋ถ€๋ถ„์— MPU6050 IMU๋ฅผ 1๊ฐœ์”ฉ, ์–‘์ชฝ ๋ฌด๋ฆŽ๋ถ€๋ถ„์— ๊ด€์ ˆ ๋‹น 3๊ฐœ์˜ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๊ฐ€ ๋ฐฐ์น˜๋˜์–ด ์žˆ๋‹ค. ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋Š” SZH-SEN01 ์ œํ’ˆ์„ ์ด์šฉํ•˜์˜€์œผ๋ฉฐ, ์ฐฉ์šฉ ๋’คํ‹€๋ฆผ ๋ฐ ์‚ฌ์šฉ์ž์— ๋”ฐ๋ฅธ ์˜ค์ฐจ๋ฅผ ์ตœ์†Œํ™”ํ•˜๊ธฐ ์œ„ํ•ด ํŒ”๊ฟˆ์น˜์™€ ํ‘ธ๋ฆ„์— ๊ฐ๊ฐ 3๊ฐœ์”ฉ ์žฅ์ฐฉํ•˜์˜€๋‹ค.

๊ฐ ์„ผ์„œ๋“ค์€ ์ƒ์ฒด๋ถ€์™€ ํ•˜์ฒด๋ถ€๋กœ ๋‚˜๋‰˜์–ด 2๊ฐœ์˜ ์•„๋‘์ด๋…ธ Due๋ณด๋“œ์— ์—ฐ๊ฒฐ์‹œ์ผฐ๋‹ค. MPU6050์€ I2C ํ†ต์‹ , ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ๋Š” ์•„๋‚ ๋กœ๊ทธ ์ž…๋ ฅ์„ ํ†ตํ•ด ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์‹ ํ•˜๋„๋ก ์ฒ˜๋ฆฌํ•˜์˜€๋‹ค. Due ๋ณด๋“œ๋Š” PC์™€์˜ serial ํ†ต์‹ ์œผ๋กœ MATLAB ํ™˜๊ฒฝ์— ์„ผ์„œ ๋ฐ์ดํ„ฐ๋ฅผ ์ „์†กํ•˜๋ฉฐ. ์‚ฌ์šฉ์ž์˜ ์›€์ง์ž„์„ ์‹ค์‹œ๊ฐ„์œผ๋กœ ๋‚˜ํƒ€๋‚ผ ์ˆ˜ ์žˆ๋„๋ก ๊ตฌ์„ฑํ•˜์˜€๋‹ค.

์‹คํ—˜ ๊ฒฐ๊ณผ ์‹œ์ž‘ ์ž์„ธ์—์„  ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜์˜ ํšจ๊ณผ๋กœ ์ธํ•ด ์˜ค์ฐจ๊ฐ€ ๊ฑฐ์˜ ์—†์Œ์„ ๊ทธ๋ฆผ 6-1๊ณผ ๊ทธ๋ฆผ 6-3์„ ํ†ตํ•ด ํ™•์ธํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค. ํ•˜์ง€๋งŒ ๋ฐ˜๋ณต๋œ ๋™์ž‘ ํ›„์—” ์ฐฉ์šฉ ์œ„์น˜์˜ ๋ณ€ํ™” ๋ฐ IMU ํŠน์„ฑ์— ์˜ํ•ด ๋“œ๋ฆฌํ”„ํŠธ๊ฐ€ ๋ฐœ์ƒํ•˜๋Š” ๊ฒƒ์„ ํ™•์ธํ•˜์˜€๋‹ค. ํŠนํžˆ ๊ทธ๋ฆผ 6-2๋Š” ๋ค๋ฒจ ๋™์ž‘์—์„œ ํŒ”์„ ์™„์ „ํžˆ ๊ตฝํžˆ์ง€ ์•Š์•˜์Œ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ์ดˆ๊ธฐ ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜ ๋œ ๊ฐ’ ๋Œ€๋น„ ๋“œ๋ฆฌํ”„ํŠธ๊ฐ€ ํฌ๊ฒŒ ๋ฐœ์ƒํ•˜์—ฌ ํŒ”์ด ์™„์ „ํžˆ ๊ตฝํ˜€์ง„ ๊ฒƒ์œผ๋กœ ๋‚˜ํƒ€๋‚˜๋Š” ๊ฒƒ์„ ๋ณผ ์ˆ˜ ์žˆ์—ˆ๋‹ค. ๋“œ๋ฆฌํ”„ํŠธ๋Š” ์šด๋™ ๋ฐ˜๋ณต ํšŸ์ˆ˜์— ๋”ฐ๋ผ ์ ์ฐจ ํฌ๊ฒŒ ๋ฐœ์ƒํ•จ์„ ํ™•์ธํ•˜์˜€์œผ๋ฉฐ, ์ด๋Š” ์—ฐ์†๋™์ž‘ ์‚ฌ์ด์— ์ •์ž์„ธ๋ฅผ ์œ ์ง€ํ•˜๋ฉด ๋‹ค์‹œ ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜์„ ํ•ด์ฃผ๋Š” ๊ณผ์ •์„ ์ถ”๊ฐ€ํ•˜์—ฌ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค. ์ด๋ฅผ ์œ„ํ•ด ๋™์ž‘๋ถ„๋ฅ˜์—์„œ nothing์ด๋ผ๋Š” class๋ฅผ ์ถ”๊ฐ€ํ•˜์˜€๋‹ค.

์ •ํ™•ํ•˜๊ฒŒ ์ž์„ธ๋ฅผ ์ทจํ•  ์ˆ˜ ์žˆ๋Š” ์ž„์˜์˜ ์‚ฌ์šฉ์ž ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•™์Šต๋œ ๋„คํŠธ์›Œํฌ๋ฅผ ์ด์šฉํ•ด ๋‹ค๋ฅธ ์‚ฌ์šฉ์ž๊ฐ€ ๊ฐ ๋™์ž‘ ๋ณ„๋กœ 20ํšŒ์”ฉ ์šด๋™ ๋™์ž‘์„ ์ˆ˜ํ–‰ํ•˜์˜€๋‹ค. ๊ทธ ๊ฒฐ๊ณผ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์ž‘์„ฑ๋œ confusion matrix๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

๊ทธ๋ฆผ. 6-1. ๋ชจ์…˜์บก์ณ plot

Fig. 6-1. Motion capture plot

../../Resources/kiee/KIEE.2023.72.11.1434/fig6.png

๊ทธ๋ฆผ. 6-2. ๋ชจ์…˜์บก์ณ plot

Fig. 6-2. Motion capture plot

../../Resources/kiee/KIEE.2023.72.11.1434/fig7.png

๊ทธ๋ฆผ. 6-3. ๋ชจ์…˜์บก์ณ plot

Fig. 6-3. Motion capture plot

../../Resources/kiee/KIEE.2023.72.11.1434/fig8.png

ํ‘œ 2. ๋™์ž‘๋ถ„๋ฅ˜ ์‹คํ—˜ ๊ฒฐ๊ณผ confusion matrix

Table 2. The confusion matrix prepared based on the results

Predict class

Nothing

Lunge

Squat

Burpee

Dumbbell

Twist

Actual

Nothing

20

0

0

0

0

0

Lunge

0

15

3

0

0

2

Squat

0

4

15

0

0

1

Burpee

0

0

0

20

0

0

Dumbbell

0

0

0

0

20

0

Twist

0

4

2

0

0

14

๊ฐ ํด๋ž˜์Šค๋งˆ๋‹ค 20๊ฐœ์˜ ๋ฐ์ดํ„ฐ๋กœ ํ…Œ์ŠคํŠธ๋ฅผ ์ง„ํ–‰ํ•œ ๊ฒฐ๊ณผ ์ •ํ™•๋„ (Accuracy) ๋ฉด์—์„œ 86.67%์˜ ์ˆ˜์น˜๋ฅผ ๋ณด์˜€์œผ๋ฉฐ f1-score๋Š” 0.869๋กœ ๊ณ„์‚ฐ๋˜์—ˆ๋‹ค. ํŠนํžˆ ์˜ค๋ถ„๋ฅ˜๊ฐ€ ๋งŽ์ด ๋œ ๋™์ž‘์€ ๋Ÿฐ์ง€์™€ ์Šค์ฟผํŠธ๋กœ ๋‚˜ํƒ€๋‚ฌ๋‹ค. ๋Ÿฐ์ง€์™€ ์Šค์ฟผํŠธ์˜ ๊ฒฝ์šฐ ์ƒ์ฒด์˜ ๋™์ž‘์€ ๋™์ผํ•˜๊ณ , ํ•˜์ฒด์˜ ๊ตฌ๋ถ€๋ฆผ๋งŒ ๋‹ค๋ฅด๊ฒŒ ๊ตฌ์„ฑ๋œ ๋™์ž‘์œผ๋กœ ๋‹ค๋ฅธ ๋™์ž‘ ๋Œ€๋น„ ์œ ์‚ฌ๋„๊ฐ€ ๋†’์€ ๋™์ž‘์ด๊ธฐ ๋•Œ๋ฌธ์— ๋ถ„๋ฅ˜ ์˜ค์ฐจ๊ฐ€ ํฐ ๊ฒƒ์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค. ํŠธ์œ„์ŠคํŠธ ๋™์ž‘์˜ ๊ฒฝ์šฐ ๋‹ค๋ฅธ ๋™์ž‘ ๋Œ€๋น„ ์ƒ์ฒด ํšŒ์ „์ด ๋งŽ์€ ๋™์ž‘์œผ๋กœ, ๋ฐ˜๋ณต๋œ ๋™์ž‘ ์‹œ ์ฐฉ์šฉํ•œ ์„ผ์„œ์˜ ๋น„ํ‹€์–ด์ง์ด ๋งŽ์ด ๋ฐœ์ƒํ•˜์—ฌ ํฐ ์˜ค์ฐจ๋ฅผ ๋‚˜ํƒ€๋‚ธ ๊ฒƒ์œผ๋กœ ํ™•์ธํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค.

4. ๊ฒฐ ๋ก 

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ์ €๊ฐ€ํ˜• IMU 6๊ฐœ์™€ ํ”Œ๋ ‰์Šค ์••๋ ฅ ์„ผ์„œ 12๊ฐœ๋ฅผ ์ด์šฉํ•ด ์‚ฌ๋žŒ์˜ ๋ชจ์…˜์บก์ฒ˜ ๋ฐ ์šด๋™ ๋™์ž‘ ๋ถ„๋ฅ˜ ๊ธฐ๋ฒ•์„ ์ œ์‹œํ•˜์˜€๋‹ค. ์ „์ด ํ•™์Šต์„ ์ด์šฉํ•˜์—ฌ ์ ์€ ๋ฐ์ดํ„ฐ์…‹์„ ๊ธฐ๋ฐ˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌ์„ฑํ•˜์˜€์Œ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ํ™œ์šฉ ๊ฐ€๋Šฅํ•œ ์ •๋„์˜ ์ •ํ™•๋„๋กœ ์‚ฌ์šฉ์ž์˜ ์šด๋™์„ ๋„์šธ ์ˆ˜ ์žˆ๋Š” ์‹œ์Šคํ…œ์„ ๊ตฌ์„ฑํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค. ํ•˜์ง€๋งŒ ๋ชจ์…˜์บก์ณ์˜ ๊ฒฝ์šฐ ๋ฐ˜๋ณต๋œ ๋™์ž‘์œผ๋กœ ์ธํ•œ ์˜ท์˜ ๋’คํ‹€๋ฆผ๊ณผ ๊ด€์„ฑ์„ผ์„œ์˜ ๋“œ๋ฆฌํ”„ํŠธ๊ฐ€ ๋ฐœ์ƒํ–ˆ์„ ๋•Œ ํšŒ๋ณตํ•  ์ˆ˜ ์—†๋Š” ์ƒํ™ฉ์ด ์ข…์ข… ๋ฐœ์ƒํ•˜์˜€๋‹ค. ์ด๋Ÿฌํ•œ ์ ๋“ค์„ ๊ฐœ์„ ํ•˜๊ธฐ ์œ„ํ•ด ํ–ฅํ›„ ์—ฐ๊ตฌ์—์„œ๋Š” ๋‹ค์–‘ํ•œ ์‚ฌ์šฉ์ž๋กœ๋ถ€ํ„ฐ ๋‹ค๋Ÿ‰์˜ ๋ฐ์ดํ„ฐ๋ฅผ ์ถ•์ ํ•ด ๋‹ค๋ฅธ ๋ฐฉ์‹์˜ ์ธ๊ณต์ง€๋Šฅ ํŒ๋ณ„ ๊ธฐ๋ฒ•์„ ๊ณ ์•ˆํ•˜์—ฌ ๋™์ž‘๋ถ„๋ฅ˜์˜ ์ •ํ™•์„ฑ์„ ๋†’์ผ ์˜ˆ์ •์ด๋‹ค. ๋˜ํ•œ ์ž์—ฐ์Šค๋Ÿฌ์šด ์›€์ง์ž„ ๋ชจ์‚ฌ๋ฅผ ์œ„ํ•ด ์‹œ๊ณ„์—ด ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ ๋”ฅ๋Ÿฌ๋‹ ๋˜๋Š” ๊ด‘ํ•™์‹ ๋ชจ์…˜์บก์ณ ๊ธฐ๋ฒ•๊ณผ์˜ ์œตํ•ฉ์„ ํ†ตํ•œ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌ์„ฑํ•  ๊ณ„ํš์ด๋‹ค.

ํ–ฅํ›„ ํ˜„์žฌ ๊ฐœ๋ฐœ ์ค‘์ธ IMU ๋Œ€๋น„ ํ›จ์”ฌ ์ €๋ ดํ•œ ํ”Œ๋ ‰์Šค ์••๋ ฅ์„ผ์„œ์™€ ๊ฒฐํ•ฉํ•˜๊ฒŒ ๋œ๋‹ค๋ฉด ์นด๋ฉ”๋ผ ์—†์ด ์‹ค๋‚ด์™ธ ๊ตฌ๋ถ„ ์—†์ด ๊ณต๊ฐ„์ œ์•ฝ์„ ๋ฐ›์ง€ ์•Š์œผ๋ฉฐ ์‚ฌ์šฉ์ž์˜ ํ˜„์žฌ ๋™์ž‘์ƒํƒœ๋ฅผ ๋ถ„๋ฅ˜ํ•˜๊ณ  ์šด๋™์„ ์ฝ”์นญํ•ด ์ค„ ์ˆ˜ ์žˆ๋Š” ์‹œ์Šคํ…œ์„ ์ €๋ ดํ•˜๊ฒŒ ๊ตฌํ˜„ํ•  ์ˆ˜ ์žˆ์„ ๊ฒƒ์œผ๋กœ ๊ธฐ๋Œ€ํ•œ๋‹ค.

Acknowledgements

์ด ๋…ผ๋ฌธ์€ ์ธ์ฒœ๋Œ€ํ•™๊ต 2022๋…„๋„ ์ž์ฒด์—ฐ๊ตฌ๋น„ ์ง€์›๊ณผ 2022๋…„๋„โ€‚์ •๋ถ€(์‚ฐ์—…ํ†ต์ƒ์ž์›๋ถ€)์˜โ€‚์žฌ์›์œผ๋กœโ€‚ํ•œ๊ตญ์‚ฐ์—…๊ธฐ์ˆ ํ‰๊ฐ€๊ด€๋ฆฌ์›์˜โ€‚์ง€์›์„ ๋ฐ›์•„โ€‚์ˆ˜ํ–‰๋œโ€‚์—ฐ๊ตฌ์ž„(No.20020741,โ€‚22kW๊ธ‰ ๊ณ ์ถœ๋ ฅ๋ฐ€๋„ LDC ํ†ตํ•ฉ ์–‘๋ฐฉํ–ฅ ์ฐจ๋Ÿ‰์šฉ ์ถฉ์ „๊ธฐ ๊ฐœ๋ฐœ)

References

1 
A. Ancans, M. Greitans, R. Cacurs, B. Banga, A. Rozentals, 2021, Wearable Sensor Clothing for Body Movement Measurement during Physical Activities in Healthcare, Sensors, Vol. 21, No. 206, pp. -DOI
2 
Sheikh MA. Iqbal, 2021, Advances in Healthcare Wearable Devices, NPJ Flexible Electronics, Vol. 5, No. 1, pp. 9-DOI
3 
H. Ceren. Ates, 2021, Wearable Devices for the Detection of COVID-19, Nature Electronics, Vol. 4, No. 1, pp. 13-14DOI
4 
Suranga. Seneviratne, 2017, A Survey of Wearable Devices and Challenges, IEEE Communications Surveys & Tutorials, Vol. 19, No. 4, pp. 2573-2620DOI
5 
Guorui. Chen, 2021, Electronic Textiles for Wearable Point-of-Care Systems, Chemical Reviews, Vol. 122, No. 3, pp. 3259-3291DOI
6 
Matteo. Menolotto, 2020, Motion Capture Technology in Industrial Applications: A Systematic Review, Sensors, Vol. 20, No. 19, pp. 5687-DOI
7 
Gutemberg. Guerra-Filho, 2005, Optical Motion Capture: Theory and Implementation, RITA, Vol. 12, No. 2, pp. 61-90DOI
8 
Alexander M. Aurand, Jonathan S. Dufour, William S. Marras, 2017, Accuracy Map of an Optical Motion Capture System with 42 or 21 Cameras in a Large Measurement Volume, Journal of Biomechanics, Vol. 58, pp. 237-240DOI
9 
Irvin Hussein. Lopez-Nava, Angelica. Munoz-Melendez, 2016, Wearable Inertial Sensors for Human Motion Analysis: A Review, IEEE Sensors Journal, Vol. 16, No. 22, pp. 7821-7834DOI
10 
Armands. Ancans, 2021, Wearable Sensor Clothing for Body Movement Measurement during Physical Activities in Healthcare, Sensors, Vol. 21, No. 6, pp. 2068-DOI
11 
Michael B. Del Rosario, Nigel H. Lovell, Stephen J. Redmond, 2016, Quaternion-Based Complementary Filter for Attitude Determination of a Smartphone, IEEE Sensors Journal, Vol. 16, No. 15, pp. 6008-6017DOI
12 
Bohan Shi, 2020, Convolutional Neural Network for Freezing of Gait Detection Leveraging the Continuous Wavelet Transform on Lower Extremities Wearable Sensors Data, 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)DOI
13 
Karthikeyan Umapathy, Sridhar Krishnan, 2006, Modified Local Discriminant Bases Algorithm and Its Application in Analysis of Human Knee Joint Vibration Signals, IEEE Transactions on Biomedical Engineering, Vol. 53, No. 3, pp. 517-523DOI
14 
Sinno Jialin Pan, Qiang Yang, 2010, A Survey on Transfer Learning, IEEE Transactions on Knowledge and Data Engineering, Vol. 22, No. 10, pp. 1345-1359DOI
15 
Christian Szegedy, 2016, Rethinking the Inception Architecture for Computer Vision, Proceedings of the IEEE Conference on Computer Vision and Pattern RecognitionDOI

์ €์ž์†Œ๊ฐœ

๋ฐฐ๋ฌธ๊ทœ (Munkyu Bae)
../../Resources/kiee/KIEE.2023.72.11.1434/au1.png

He received the B.S degree in Electrical Engineering from Incheon National University, South Korea in 2022.

Currently, he is pursing M.S degree in the same institution.

His research interest include machine learning, optimal control, autonomous vehicle, SLAM, 3D perception.

๊น€๊ฑดํƒœ (Guntae Kim)
../../Resources/kiee/KIEE.2023.72.11.1434/au2.png

He received the B.S degree in Electrical Engineering from Incheon National University, South Korea in 2023.

Currently, he is pursing the combined Masters and PhD degree in the same institution.

His research interest include optimal control, linear control, data driven modeling, autonomous vehicle, SLAM, path planning.

๋ฐ•์ดํ˜• (Yi-Hyeong Park)
../../Resources/kiee/KIEE.2023.72.11.1434/au3.png

He received the B.S degree in Computer Science from Bachelor's Degree Examination for Self-Education.

Currently, he is pursuing the B.S degree in Embedded System engineering from Incheon National University.

His research interest include intelligent automation, autonomous robotics, signal pattern recognition.

๊ฐ•์ฐฝ๋ฌต (Chang Mook Kang)
../../Resources/kiee/KIEE.2023.72.11.1434/au4.png

He received the B.D and Ph.D degrees in Electrical Engineering from Hanyang University, Seoul, South Korea, in 2012 and 2018 respectively.

He was a Senior Engineer with Agency for Defense Development, Daejeon, Korea, from 2018 to 2019.

He is currently an Associate Professor with the Department of Electric Engineering, Incheon National University, Incheon South Korea.

His research interest include linear system, optimal control, autonomous vehicle and artificial intelligent.