ドライバ運転感覚の定量化に向けた筋電位特徴量の抽出

全文

(1)Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. υϥΠόӡస‫֮ײ‬ͷఆྔԽʹ޲͚ͨ‫ిے‬Ґಛ௃ྔͷநग़ ௗ‫ߞ ډ‬ଠ1,a). Տத ࣏थ1,b). ௕੉ ޭࣇ2,c). খ‫࣍޺ ܀‬1,d). ֓ཁɿࣗಈंͷ։ൃʹ͓͍ͯӡస‫֮ײ‬ΛఆྔతʹධՁ͢ΔͨΊʹɼυϥΠό͕஌֮͢Δं྆ಛੑͷมԽͱ ૬ؔͷߴ͍ੜମ৴߸Λநग़͢ΔɽυϥΠϏϯάγϛϡϨʔλΛ༻͍ͯ௨ৗं྆ͱΤϯδϯ-λΠϠؒͷΪΞ ൺΛେ͖ͨ͘͠ं྆ͷ 2 छྨͰ௥ै૸ߦΛߦͬͨɽυϥΠόͷ‫ిے‬ਤɼࢹઢ৘ใɼൽෘ৘ใΛଌఆ͠ɼं ؒ‫཭ڑ‬ௐઅதͷಛ௃ྔΛநग़ͨ͠ɽk-means ๏ʹΑΓੜମ৴߸Λ෼ྨͨ݁͠Ռɼं྆ಛੑͷҧ͍ʹΑͬͯ එɼ๹ɼ࿹ͷ‫ిے‬Ґʹҧ͍͕‫ݱ‬ΕΔ͜ͱΛࣔͨ͠ɽ. Extraction of Electromyogram Features for Quantification of Driving Sensation Kota Torii1,a). Haruki Kawanaka1,b). 1. ͸͡Ίʹ. Koji Nagase2,c). Koji Oguri1,d). มԽͳͲɼੜମ৴߸Λ༻͍ͨ٬‫؍‬తධՁํ๏͕ఏҊ͞Εͯ ͍Δ [4], [5], [6], [7]ɽ͜ΕΒͷઌߦ‫ڀݚ‬͸ಛੑͷ͕ࠩग़΍. ࣗಈंͷӡస‫֮ײ‬ධՁʹ͸ɼं྆ͷৼಈΛ௚઀‫ܭ‬ଌ͢Δ. ͍͢ટճ࣌ɾंઢมߋ࣌ʹண໨͍ͯ͠Δɽ͔͠͠ɼ࣮ࡍͷ. ධՁํ๏ [1] ͱɼख़࿅υϥΠόʹΑΔ‫׭‬ೳධՁΛ༻͍ΔධՁ. ૸ߦʹ͓͍ͯ͸ɼટճ࣌΍ंઢมߋ࣌ʹൺ΂ͯ௚ઢ࿏૸ߦ. ํ๏ [2], [3] ͕͋Δɽલऀ͸ं྆ʹൃੜ͢Δৼಈ΍ࣗं଎౓. ΍ઌߦं௥ै૸ߦ͕ओʹͳΔ͜ͱ͕༧૝͞ΕΔɽͦ͜Ͱɼ. ੒෼ͳͲ͔ΒධՁ͢Δɽ‫ऀޙ‬͸ɼ࣮ࡍʹਓ͕ࣗಈंʹ৐ͬ. ຊ‫Ͱڀݚ‬͸࣮ࡍͷ૸ߦͰத৺ͱͳΔͱߟ͑ΒΕΔ௥ै૸ߦ. ͨࡍͷ‫ײ‬ੑΛ‫ܗ‬༰ࢺ΍‫׭‬ೳධՁʹ‫ͮ͘ج‬ධՁࣜΛ༻͍ͯධ. ࣌ʹ͍ͯಛੑͷҟͳΔं྆Λ૸ߦͨ͠ࡍʹɼಛੑʹ௥ै͠. Ձ͢Δɽ͔͠͠ɼ͜ΕΒͷ‫׭‬ೳධՁ͸͍ͣΕ΋ओ‫؍‬తͳධ. ͯมԽ͢Δੜମ৴߸ಛ௃ྔΛநग़͢Δɽ௥ै૸ߦ࣌ʹҧ͍. Ձͱͳ͍ͬͯΔɽͦͷͨΊं྆ͷ໰୊఺ʹ͍ͭͯͲͷఔ౓. ͕දΕΔಛੑͱͯ͠ΤϯδϯʖλΠϠؒͷΪΞൺʢΪΞൺ. վળ͢΂͖͔͕ෆ໌ྎͰɼվળલͱվળ‫Ͱޙ‬ධՁΛߦͬͯ. ಛੑʣʹண໨͠ɼυϥΠϏϯάγϛϡϨʔλͷΪΞൺಛੑ. ΋ҟͳΔҙ‫͕ݟ‬ग़Δ৔߹͕͋Δɽ͕ͨͬͯ͠υϥΠόͷओ. Λ 2 छྨʹมߋͯ͠‫܁‬Γฦ͠‫ܭ‬ଌ͠܏޲Λ෼ੳ͢Δɽ. ‫؍‬Λ٬‫؍‬తʹ਺஋ͱͯ͠஌Δ͜ͱͷͰ͖Δ௥Ճͷࢦඪ͕ඞ ཁͱ͞Ε͍ͯΔɽ ख़࿅υϥΠόͷओ‫؍‬஋ʹΑΔධՁΛɼఆྔతͳ஋ʹΑΓ. 2. ੜମ৴߸ ௥ै૸ߦ࣌ʹΪΞൺಛੑ͕ҟͳ͍ͬͯΔͱυϥΠό͕ͦ. ิॿ͢Δ͜ͱͰɼख़࿅υϥΠό͕ΑΓਖ਼֬ͳ਍அΛ͢Δ͜. ͷҧ͍Λ‫ʹࡍͨ͡ײ‬ෆշ‫ײ‬΍ෆ҆‫͕ײ‬ੜ͡Δ͜ͱ͕͋Δɽ. ͱ͕Ͱ͖ΔΑ͏ʹͳΔɽ͜Ε·Ͱʹɼ࿹΍‫إ‬ද৘ͷ‫ిے‬Ґ. Αͬͯຊ‫Ͱڀݚ‬͸ɼ࣍ʹ‫͛ڍ‬Δઌߦ‫Ͱڀݚ‬ෆշɼෆ͓҆Α. 1. ͼ‫ۓ‬ுʹؔ࿈͕͋Δͱ͞Ε͍ͯΔ‫ిے‬Ґɼࢹઢ৘ใ͓Αͼ. 2. a) b) c) d). Ѫ஌‫ཱݝ‬େֶ େֶӃ৘ใՊֶ‫ڀݚ‬Պ Graduate School of Information Science and Technology, Aichi Prefectural University, Nagakute, Aichi 480-1198, Japan τϤλςΫχΧϧσΟϕϩοϓϝϯτ‫ࣜג‬ձࣾ Toyota Technical Development Corporation, Toyota, Aichi 470-0334, Japan [email protected] [email protected] [email protected] [email protected]. c 2017 Information Processing Society of Japan . ൽෘ৘ใΛ‫ܭ‬ଌ͠ɼղੳʹ༻͍Δɽ. 2.1 ‫ిے‬Ґ ‫ిے‬Ґͱ͸ɼ‫ے‬೑Λಈ͔ͨ͢Ίʹੜ͡ΔిҐͷ͜ͱͰ ͋Δɽ‫ے‬ͷ์ిྔ͓Αͼ์ిͷλΠϛϯά͔Βɼ‫ے‬೑ͷ ‫׆‬ಈҐஔ΍‫׆‬ಈ࣌ؒɼස౓Λ೺ѲͰ͖ΔɽFig. 1ɼFig. 2 ʹͦΕͧΕ‫إ‬໘ɼ࿹ͷ‫ے‬ͷҐஔΛࣔ͢ɽӡస࣌ʹෆ҆΍. 1.

(2) Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. F (t) =. RMS(t) %MVC. (2). ·ͨɼ‫ిے‬Ґͷप೾਺੒෼͸‫ے‬ർ࿑Λࣔ͢ಛ௃ྔͩͱ͍Θ. Corrugator supercilii. Ε͓ͯΓ [9]ɽप೾਺੒෼Λಋग़ͯ͠ಛ௃తͳ਺஋Λ‫ࢉܭ‬ ͢Δ͜ͱͰ‫ے‬ർ࿑ͳͲΛ‫ݕ‬ग़͢Δ͜ͱ͕Ͱ͖ΔɽҰൠతʹ. Zygomaticus major. ‫ے‬ણҡ্Λి‫ؾ‬త‫ڵ‬ฃ͕఻೻͢Δࡍͷ఻ಋ଎౓͸୅ँ࢈෺ ͷ஝ੵʹΑͬͯ௿Լ͢ΔͨΊɼ‫͕ے‬ർ࿑͢Δʹ͕ͨͬͯ͠. Masseter. प೾਺͕௿஋ʹͳΔ͜ͱ͕஌ΒΕ͍ͯΔɽΤϯδϯτϧΫ ͷҟͳΔं྆Λӡసͨ͠ࡍʹ͸଎౓ௐ੔͕ࠔ೉ʹͳΓർ࿑. ਤ 1: ಛ௃ྔͱͯ͠༻͍Δ‫إ‬໘‫ిے‬෦Ґ. Λ༠ൃ͢Δͱߟ͑ΒΕΔɽΑͬͯɼ‫ిے‬Ґͷप೾਺੒෼ʹ. Fig. 1 Face parts used as electromyogram features. Τϯδϯಛੑͷҧ͍͕දΕΔͱ༧૝͞ΕΔɽ۩ମతʹ͸ղ ੳ۠ؒʹର͠ߴ଎ϑʔϦΤม‫( ׵‬Fast Fourier Transform:. FFT) Λߦͬͨࡍͷதԝप೾਺ (MeDian power Frequency :MDF) ΍ɼୈ 1 ϐʔΫप೾਺͓Αͼୈ 2 ϐʔΫप೾਺Λͱ. Brachioradialis muscle. ΔɽMDF ͷఆٛΛࣜ (3) ʹࣔ͢ɽप೾਺ f ຖͷৼ෯஋Λ. P (f ) ͱ͢Δɽ‫ిے‬ҐͷαϯϓϦϯάप೾਺Λ F s ͱ͢Δɽ. . . MDF. Fs 2. P (f )df =. P (f )df. (3). MDF. 0. 2.2 ࢹઢ৘ใ ࢹઢ৘ใͱͯ͠ɼॠ໨਺΍ಏ޸‫ܘ‬ɼαοΧʔυɼ։‫཰؟‬ Λ‫ܭ‬ଌ͢Δɽॠ໨਺͸୯Ґ࣌ؒ͋ͨΓͷॠ͖ͷճ਺Λද ਤ 2: ಛ௃ྔͱͯ͠༻͍Δ࿹‫ిے‬෦Ґ. ͢ɽ‫ܭ‬ଌ࣌ؒ l ඵʹ͓͚Δ߹‫ܭ‬ͷॠ͖ճ਺Λ b ճͱͯ͠ɼ. Fig. 2 Arm part used as electromyogram features. ࣜ (4) ʹࣔ͢ํ๏Ͱࢉग़͢Δɽ. ෆշΛ‫͡ײ‬Δ৔߹ʹ͸ Fig. 2 ͷલ࿹‫ے‬ͷऩॖྗʹ͕ࠩੜ. br =. ͡ɼ‫ۓ‬ு͕දΕΔ [7]ɽ·ͨɼFig. 1 ʹࣔ͢Α͏ͳᚧඑ‫ے‬ɼ. b l. (4). େ๹ࠎ‫ے‬ɼᄐ‫ے‬Λ࢝Ίͱ͢Δද৘‫ʹے‬͸ෆշͷ৘ಈ͕ද. ॠ໨਺͸ूதྗΛཁ͢Δ৔໘Ͱ‫ݮ‬গ͠ɼෆ҆ɾ‫ڵ‬ฃঢ়ଶͰ. ग़͢Δ [8]ɽ ΪΞൺಛੑ͕௨ৗͱҟͳΔं྆ʹ৐ंͨ͠ࡍ. ͸૿Ճ͢Δ܏޲ʹ͋Δ [10]ɽ·ͨɼ֮੧౒ྗΛ͍ͯ͠Δঢ়. ʹ͸ɼෆ҆΍‫ۓ‬ு͕ੜ͡‫ిے‬Ґͷৼ෯͕େ͖͘ͳΔͱ༧. ଶͰ͸ॠ໨਺͕૿Ճ͢Δ [11]ɽΪΞൺಛੑͷҟͳΔं྆ʹ. ૝͞ΕΔɽຊ‫Ͱڀݚ‬͸‫ిے‬Ґ͸ɼඑؒ෇ۙͷ‫ے‬೑Ͱ͋Δ. ৐ंͨ͠ࡍ͸ΑΓણࡉͳૢ࡞͕‫ٻ‬ΊΒΕΔͨΊूதྗ͕ߴ. ᚧඑ‫( ے‬Corrugator supercilii)ɼ๹ͷ‫ے‬೑Ͱ͋Δେ๹ࠎ‫ے‬. ·ͬͨΓɼΪΞൺಛੑͷࠩΛҙࣝ͠߹ΘͤΑ͏ͱ͢Δ͜ͱ. (Zygomaticus major)ɼֺͷ‫ے‬೑Ͱ͋Δᄐ‫( ے‬Masseter)ɼ࿹. ʹΑͬͯॠ໨਺͕มԽͨ͠Γ͢Δͱ༧૝͞ΕΔɽ. ͷ‫ے‬೑Ͱ͋Δ࿹ᒷࠎ‫( ے‬Brachioradialis muscle) ͷ 4 Χॴ. αοΧʔυ͸୹࣌ؒͷߴ଎ͳ‫ٿ؟‬ӡಈͰ͋Γɼࢹ໺पล ͷର৅ʹ໨Λ޲͚Δ࣌ͷ͓Αͦ 50 ms Ҏ಺ͷ‫ٸ‬଎ͳҠಈͷ. Ͱ‫ܭ‬ଌ͢Δɽ Ұൠతͳडಈి‫ۃ‬Λ༻͍Δ৔߹ɼి‫ۃ‬த৺‫཭ڑ‬Λ 10 mm ͔Β 30 mm ʹ͢Δɽ‫ిے‬Ґ৴߸ͷฏ‫ৼۉ‬෯Λಛ௃ྔͱ. ͜ͱΛࢦ͢ɽαοΧʔυճ਺ sr ͸‫ܭ‬ଌ࣌ؒ l ඵʹ͓͚Δ ߹‫ܭ‬ͷճ਺Λ s ճͱͯ͠ɼࣜ (5) Λ༻͍ͯࢉग़͢Δɽ. ͢ΔͨΊɼࣜ (1) ʹࣔ͞ΕΔೋ৐ฏ‫ۉ‬ฏํࠜ (Root Mean. Square :RMS) ΛͱΔɽ T 1 RMS(t) = e2 (t + Υ)dΥ 2T −T. sr =. s l. (5). αοΧʔυ͸஫໨͢Δର৅ʹରͯ͠໨Λ޲͚Δ࣌ͷಈ࡞Ͱ. (1). ͋Δ͕ɼ஫໨͢Δඞཁ͕͋Δର৅͕૿Ճ͢Ε͹αοΧʔυ ͷճ਺΋૿Ճ͢Δ [12]ɽΪΞൺͷҟͳΔं྆Λӡసͨ͠ࡍ. ͜͜Ͱɼe(t) ͸‫ిے‬Ґ৴߸Ͱɼ(−T, T ) ͕‫ͳʹؒ۠ࢉܭ‬Δɽ. ʹ͸஫ҙྗ͕ߴ·Γ௨ৗΑΓ΋αοΧʔυͷճ਺͕૿Ճ͢. ‫ݸ‬ਓؒɼ‫Ͱؒے‬ൽԼࢷ๱ͷް͞ɼൽෘΠϯϐʔμϯε͕ҟ. Δͱ༧૝͞ΕΔɽ. ͳΔͨΊɼԾʹ‫ے‬ઢҡϨϕϧͰൃੜ͍ͯ͠ΔిҐ͕ಉ͡Ͱ. ಏ޸‫ܘ‬͸೒࠼ͷ্୺͔ΒԼ୺·Ͱͷ௕͞ͷ͜ͱͰ͋Δɽ. ͋ͬͯ΋ɼి‫ۃ‬ϨϕϧͰ‫ه‬࿥͞ΕΔిҐ͸ҟͳΔɽͦͷͨ. ಏ޸‫ܘ‬ͷมԽ͸໢ບʹ౤ࣹ͢Δޫྔͷௐ੔͕ओͳཁҼͱ. Ίࣜ (2) ʹࣔ͢Α͏ʹ࠷େਵҙऩॖྗ (Maximal Voluntary. ͞Ε͍ͯΔ͕ɼͦͷଞʹࣗ཯ਆ‫ܥܦ‬ͷࢧ഑Λड͚ͯมԽ͢. Contraction :MVC) Ͱ RMS Λআ͢Δ͜ͱʹΑΓਖ਼‫ن‬ԽΛ. Δ [13]ɽަ‫ײ‬ਆ‫ܦ‬༏ҐʹͳΔͱಏ޸‫ܘ‬͸େ͖͘ͳΓɼ෭ަ. ߦ͏ɽ. ‫ײ‬ਆ‫ܦ‬༏ҐʹͳΔͱಏ޸‫ܘ‬͸খ͘͞ͳΔɽΪΞൺ͕ҟͳΔ. c 2017 Information Processing Society of Japan . 2.

(3) Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ं྆Λӡసͨ͠ࡍʹ͸ަ‫ײ‬ਆ‫͕ܥܦ‬ဏਐ͠ɼಏ޸‫͕ܘ‬େ͖. ද 1: ಛ௃ྔҰཡ. ͘ͱ༧૝͞ΕΔɽӡస։࢝࣌ͷಏ޸‫ʹܘ‬ର͢Δऴྃ࣌ I ඵ ͷಏ޸‫ܘ‬ͷ௕͞ p(I) Λಛ௃ྔͱ͠ɼࣜ (6) ʹࣔ͢ɽ. Table 1 Feature list ൪߸. ಛ௃ྔ໊. ൪߸. ಛ௃ྔ໊. 1. ᚧඑ‫ ࠨے‬Ave. 24. ᄐ‫ے‬ӈ SD. 2. ᚧඑ‫ے‬ӈ Ave. 25. ᄐ‫ ࠨے‬MDF. 3. ᚧඑ‫ ࠨے‬SD. 26. ᄐ‫ے‬ӈ MDF. ։‫཰؟‬͸ಏ޸͕ද໘ʹͲΕ͚ͩ࿐ग़͍ͯ͠Δ͔Λඦ෼཰Ͱ. 4. ᚧඑ‫ے‬ӈ SD. 27. ᄐ‫ ࠨے‬1st Freq. දͨ͠΋ͷͰ͋Γɼࣜ (7) Λ༻͍ͯղੳ۠ؒ։࢝࣌ͷ։‫؟‬. 5. ᚧඑ‫ ࠨے‬MDF. 28. ᄐ‫ے‬ӈ 1st Freq. ཰͔Βղੳ۠ؒऴྃ࣌ͷ։‫཰؟‬ͷมԽ཰Λղੳ͢Δɽ͜͜. 6. ᚧඑ‫ے‬ӈ MDF. 29. ᄐ‫ ࠨے‬2nd Freq. Ͱɼ։‫཰؟‬͸‫ܭ‬ଌ࣌ؒ l ඵؒʹ͓͚Δ։‫཰؟‬Λ e(l) ͱ͢Δɽ. 7. ᚧඑ‫ ࠨے‬1st Freq. 30. ᄐ‫ے‬ӈ 2nd Freq. 8. ᚧඑ‫ے‬ӈ 1st Freq. 31. ࿹ᒷࠎ‫ ࠨے‬Ave. 9. ᚧඑ‫ ࠨے‬2nd Freq. 32. ࿹ᒷࠎ‫ے‬ӈ Ave. 10. ᚧඑ‫ے‬ӈ 2nd Freq. 33. ࿹ᒷࠎ‫ ࠨے‬SD. 11. େ๹ࠎ‫ ࠨے‬Ave. 34. ࿹ᒷࠎ‫ے‬ӈ SD. 12. େ๹ࠎ‫ے‬ӈ Ave. 35. ࿹ᒷࠎ‫ ࠨے‬MDF. 13. େ๹ࠎ‫ ࠨے‬SD. 36. ࿹ᒷࠎ‫ے‬ӈ MDF. ҟͳΔं྆Λӡసͨ͠ࡍʹ͸ɼӡస‫֮ײ‬ͷҧ͍ʹΑΔ֮੧. 14. େ๹ࠎ‫ے‬ӈ SD. 37. ࿹ᒷࠎ‫ ࠨے‬1st Freq. ޮՌ͔Β։‫্͕཰؟‬ঢ͢Δͱ༧૝͞ΕΔɽ. 15. େ๹ࠎ‫ ࠨے‬MDF. 38. ࿹ᒷࠎ‫ے‬ӈ 1st Freq. 16. େ๹ࠎ‫ے‬ӈ MDF. 39. ࿹ᒷࠎ‫ ࠨے‬2nd Freq. 17. େ๹ࠎ‫ ࠨے‬1st Freq. 40. ࿹ᒷࠎ‫ے‬ӈ 2nd Freq. 18. େ๹ࠎ‫ے‬ӈ 1st Freq. 41. ඓ෦ൽෘԹ౓ ૿Ճྔ. 19. େ๹ࠎ‫ ࠨے‬2nd Freq. 42. ൽෘిҐ ૿Ճྔ. 20. େ๹ࠎ‫ے‬ӈ 2nd Freq. 43. ॠ໨਺. ΔɽൽෘԹ౓ΛଌΔ͜ͱͰɼͦͷ෦Ґͷໟࡉ݂؅ʹ͓͚Δ. 21. ᄐ‫ ࠨے‬Ave. 44. ಏ޸‫ ܘ‬Ave. ݂ྲྀྔ͕Θ͔Δɽ‫إ‬ද໘ͷൽෘ౓Թ͸ަ‫ײ‬ਆ‫ʹܥܦ‬ΑΔਆ. 22. ᄐ‫ے‬ӈ Ave. 45. αοΧʔυճ਺. ‫ࢧܦ‬഑Λड͚Δ [15]ɽͦͷͨΊɼަ‫ײ‬ਆ‫ܥܦ‬ͷ݂؅ऩॖ࡞. 23. ᄐ‫ ࠨے‬SD. 46. ։‫ ཰؟‬Ave. pr =. p(l) p(1). e(l) er = e(1). (6). (7). ֮੧ਫ४Λࣔ͢ࢦඪͱͯ͠ར༻͞Ε͓ͯΓɼӡస‫཰֮ײ‬͸ ൓Ԡ࣌ؒͱ૬͕ؔ͋Δ͜ͱ͕ࣔ͞Ε͍ͯΔ [14]ɽΪΞൺͷ. 2.3 ൽෘ৘ใ ൽෘ৘ใͱͯ͠ඓ෦ൽෘԹ౓͓ΑͼࢦઌͷൽෘిҐΛ‫ܭ‬ ଌ͢ΔɽൽෘԹ౓ͷ‫ܭ‬ଌʹ͸઀৮‫ܕ‬ͷԹ౓‫Ͱܭ‬௚઀‫ܭ‬ଌ͢. ༻ʹΑΓൽෘԹ౓͸௿Լ͠ɼަ‫ײ‬ਆ‫׆ܦ‬ಈͷ཈੍ʹΑΓ݂ ྲྀੑ͕૿Ճ͠ൽෘԹ౓্͕ঢ͢Δɽಛʹඓ෦͸ಈ຺ͱ੩຺. Λࣝผ͢Δͷʹ༗ޮͳಛ௃ྔΛ k-means ๏Λ༻͍ͯநग़. ͕ަΘΔ෦෼Ͱ͋ΔͨΊԹ౓มԽ͕‫ݦ‬ஶʹग़΍͍͢ [15]ɽ. ͢Δɽ. ΪΞൺͷҧ͍͔ΒετϨε͕ੜ͡ɼަ‫ײ‬ਆ‫ܹ͕ࢗ͞ܥܦ‬Εɼ. ͦΕͧΕͷ෦Ґ͔ΒಘΒΕΔ৴߸͔ΒҰఆ۠ؒͷฏ‫ۉ‬஋. ൽෘԹ౓͕௿Լ͢Δͱ༧૝͞ΕΔɽ‫ܭ‬ଌ࣌ؒ l ඵؒʹ͓͚. (Ave)ɼඪ४ภࠩ (SD)ɼதԝप೾਺ (MDF)ɼϐʔΫप೾਺. ΔൽෘԹ౓Λ st (l) ͱͯࣜ͠ (8) Λ༻͍ͯऴྃ࣌ͷมԽΛ. (Freq) ͳͲΛ౷‫ͯ͠ͱྔܭ‬நग़͠ɼTable 1 ʹࣔ͢‫ ܭ‬46 ‫ݸ‬. ղੳ͢Δɽ. st (l) st = st (1). (8). ΨϧόχοΫൽෘ൓Ԡ (Galvanic Skin Response :GSR) ͸ɼ. ͷಛ௃ྔ‫܈‬ΛಘΔɽ͜ΕΒͷಛ௃ྔʹ͸࣮ࡍʹ͸ं྆ಛੑ ͱ͸ؔ܎ͷͳ͍ɼෆཁͳಛ௃ྔ͕‫·ؚ‬Ε͍ͯΔͱߟ͑ΒΕ Δɽ͜ͷಛ௃‫͔܈‬Β m(1 < m < 6) ‫ݸ‬બ୒ͯ͠ m ࣍‫ݩ‬ͷ ಛ௃ϕΫτϧ xi (1 < i < 46 Cm ) Λੜ੒͢Δɽ͜͜Ͱ i ͸ಛ. ൃ‫ʹ׼‬ΑΔίϯμΫλϯεͷ૿ՃΛଌఆ͢Δ΋ͷͰ͋Γɼ. ௃ྔ൪߸Λࣔ͢ɽ࣮‫ݧ‬ͷηοτ਺Λ n ηοτͱ͠ɼn ࢼߦ. ަ‫ײ‬ਆ‫ʹܥܦ‬Αܹͬͯࢗ͞Εͨ‫׼‬થͷ‫׆‬ੑมԽʹؔ࿈͕ਂ. i = {xi1 , xi2 , · · · , xin } ʹ k-means ͷಛ௃ϕΫτϧू߹ X. ͍ [16]ɽ৺ཧతಈ༳͔Βަ‫ײ‬ਆ‫ܹ͕ࢗ͞ܥܦ‬Εͯग़ͨྫྷ΍. ๏Λద༻͢Δɽ. ‫׼‬ΛɼίϯμΫλϯεͷ্ঢʹΑΓ‫ݕ‬஌͢Δɽྫྷ΍‫Ͳͳ׼‬. k-means ๏Ͱ͸ɼΫϥελ j ͷॏ৺ cij ʹର͢Δࣜ (10). ͷਫ਼ਆੑൃ‫׼‬͸ɼखͷͻΒɼࢦͷෲଆɼ଍ͷཪʹ‫ݶ‬ఆ͞Ε. ͷධՁؔ਺ g Λ࠷খԽ͢ΔΑ͏ʹ k ‫ݸ‬ͷΫϥελʹ෼ׂ. ͍ͯΔɽͦͷͨΊɼຊ‫Ͱڀݚ‬͸ਓࠩ͠ࢦʹి‫ۃ‬Λ‫͚͖ͭר‬. ͢Δɽ. ίϯμΫλϯε্ঢΛ‫ݕ‬ग़͢ΔɽΪΞൺಛੑ͕ҟͳΔं྆ ʹ৐ͬͨࡍʹਫ਼ਆෛՙ͔Βྫྷ΍‫׼‬Λ͔͖ɼίϯμΫλϯε ্͕ঢ͢Δͱ༧૝͞ΕΔɽ‫ܭ‬ଌ࣌ؒ l ඵʹ͓͚ΔൽෘిҐ Λ G(l) ͱͯࣜ͠ (9) Λ༻͍ͯऴྃ࣌ͷมԽΛղੳ͢Δɽ. G=. G(l) G(1). (9). 3. ಛ௃ྔநग़ख๏. g=. k . ||xi − cij ||2. (10). j=1 xi ∈xi. ͜ͷΑ͏ͳΫϥελϦϯάΛ‫܁‬Γฦͯ͠ߦ͏ɽͨͩ͠ɼຊ ‫Ͱڀݚ‬͸ΪΞൺಛੑ͕௨ৗͲ͓Γͷं྆ͱҟৗͳं྆ͱͰ ଌఆΛߦ͏ͷͰ k = 2 ͱͨ͠ɽͦͷ‫ޙ‬ɼΫϥελͷੜ੒݁ Ռ͕ΪΞൺಛੑͷਖ਼ɾҟৗͱ߹க͍ͯ͠Δׂ߹Λਖ਼ղ཰ͱ ͯ͠ࢉग़͢Δɽͦͯ͠ɼಛ௃ϕΫτϧͷ࣍‫ ਺ݩ‬m ͝ͱʹ ࠷΋ਖ਼ղ཰ͷߴ͍ಛ௃ͷ૊Έ߹Θͤ i Λ‫ٻ‬ΊΔɽ. લઅͰड़΂༷ͨʑͳੜମ৴߸ͷಛ௃ྔ͔ΒɼΪΞൺಛੑ. c 2017 Information Processing Society of Japan . 3.

(4) Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. Pre car start. Analysis section. Driver’s car start. 10 s. Cruise (60 km/h). Pre car Driver. Ready to start. Driving (Up to 80 km/h). Approach. Follow Pre car. ਤ 5: ࣮‫ݧ‬ϓϩτίϧ Fig. 5 Experiment protocol. ਤ 3: υϥΠόࢹ఺ Accuracy rate(0~1). Fig. 3 Driver’s view. 0.80 0.75 0.70 0.65. 0.74. 0.74. 0.75. 0.73 0.70. 0.69. 0.60 1. 2. 3. 4. 5. 6. 7. Vector dimensions. ਤ 4: ၆ᛌࢹ఺. ਤ 6: ඃ‫ڞऀݧ‬௨ͷ༗ޮಛ௃ྔΛద༻ͨ͠৔߹ͷਖ਼ղ཰. Fig. 4 Bird’s eye view. Fig. 6 Accuracy rate by common features for all subjects. 4. ࣮‫ݧ‬. Λ༻͍ͨ৔߹ͷਖ਼ղ཰ͷฏ‫ۉ‬஋Λ͍ࣔͯ͠Δɽಛ௃ྔ਺ 1 Ͱબ͹Εͨ΋ͷ͸එͷ‫ిے‬Ґͷप೾਺੒෼Ͱ͋ͬͨɽਖ਼ղ. ΪΞൺಛੑͷมԽΛࣝผͰ͖Δੜମ৴߸Λநग़͢ΔͨΊ. ཰͕ 0.7 Λ௒͔͑ͯΒਖ਼ղ཰͕Լ͕Γ࢝ΊΔ·Ͱͷಛ௃ྔ. ʹυϥΠϏϯάγϛϡϨʔλΛ༻͍ͯ‫ܭ‬ଌ࣮‫ݧ‬Λߦͬͨɽ. ਺ 2 ͔Β 4 ·ͰͷਪఆͰ࢖༻͞Εͨಛ௃ྔͱɼಛ௃ྔ਺ 1. ࢹઢ‫ܭ‬ଌʹ͸ Seeing Machine ࣾͷ Facelab γεςϜΛ࢖. Ͱબ͹Εͨಛ௃ྔͱͦΕΒͷճ਺Λ Table 2 ʹࣔ͢ɽ එɼ. ༻͢Δɽ. ๹ɼ࿹ͷ‫ిے‬Ґ͕ΪΞൺಛੑͷࣝผʹ༗ޮͰ͋Δ͜ͱ͕Θ ͔ͬͨɽ. 4.1 ࣮‫ݧ‬ϓϩτίϧ. ࠓճ‫ܭ‬ଌͨ͠එɼ๹ɼ͋͝ɼ࿹ͷ‫ిے‬Ґͷৼ෯೾‫ͭʹܗ‬. ‫཭ڑ‬໿ 1200 m ͷ௚ઢίʔεΛ૸ߦ͢ΔɽFig. 3 ͓Αͼ. ͍ͯ Fig. 7 ʹࣔ͢ɽॎ͕࣠ 100%MVC ʹΑΓਖ਼‫ن‬Խ͞Ε. Fig. 4 ʹઌߦं௥ै૸ߦ࣌ͷυϥΠόࢹ఺ͱ၆ᛌࢹ఺Λࣔ. ͨ‫ిے‬Ґͷ‫͞ڧ‬Λද͓ͯ͠Γɼԣ͕࣠ղੳ۠ؒͷ࣌‫ྻܥ‬Λ. ͢ɽ࣮‫ݧ‬ϓϩτίϧΛ Fig. 5 ʹࣔ͢ɽઌߦं͕૸ߦΛ։. ද͢ɽ੨৭ͷ೾‫ܗ‬͸௨ৗं྆ͷ૸ߦʹ͓͚Δର৅‫ిے‬Ґͷ. ͔࢝ͯ͠Β 10 ඵ‫ंࣗʹޙ‬Λൃਐͤ͞Δɽͦͷ‫ޙ‬ɼ੍‫ݶ‬଎. 10 ηοτ෼ͷதԝ஋Λද͓ͯ͠Γɼ੺৭ͷ೾‫ܗ‬͸ΪΞൺಛ. ౓Λ௒͑ͳ͍Α͏ʹ͠ͳ͕Βઌߦंʹ઀ۙ͠ɼυϥΠόͷ. ੑͷେ͖ͳं྆ͷ૸ߦʹ͓͚Δର৅‫ిے‬Ґͷ 10 ηοτ෼. ೚ҙͷंؒ‫Ͱ཭ڑ‬௥ै૸ߦΛߦ͏ɽFig. 5 ʹࣔ͢Α͏ͳɼ. தԝ஋Λද͍ͯ͠Δɽk-means ๏ʹΑΓબ͹Εͨಛ௃ྔͰ. ‫ݮ‬଎։͔࢝ΒυϥΠό͕௥ै૸ߦʹࢸΔ·Ͱͷ۠ؒΛղੳ. ͋Δɼඑɼ๹ɼ࿹ͷ 3 छྨʹؔͯ͠ΪΞൺಛੑͷҟͳΔं. ۠ؒͱͨ͠ɽΪΞൺಛੑͷਖ਼ৗͳं྆Ͱ 1 ճ૸ߦͨ͠‫ʹޙ‬. ྆ʹ৐ͬͨࡍͷ೾‫͕ܗ‬௨ৗͱൺֱͯ͠େ͖͘ͳΔ͜ͱ͕֬. ΪΞൺಛੑ͕ҟͳΔं྆Λ 1 ճ૸ߦ͢ΔͷΛ 1 ηοτͱ. ೝͰ͖ͨɽ. ͠ɼඃ‫ऀݧ‬ຖʹ 10 ηοτ (n = 20) ૸ߦͨ͠ɽඃ‫ऀݧ‬͸ 3. ࣍ʹɼඃ‫࡯ߟʹͱ͝ऀݧ‬Λߦ͏ɽඃ‫ऀݧ‬ຖͷਖ਼ղ཰ͷਪ. ໊Ͱීஈ͔Β௨ֶ΍௨‫Ͱۈ‬ӡసΛ͍ͯ͠Δɽ‫ܭ‬ଌ͞Εͨੜ. ҠΛ Fig. 8 ʹࣔ͢ɽ ඃ‫ ऀݧ‬A Ͱ͸එɼ࿹ͷ‫ిے‬Ґ͕બ. ମ৴߸ʹରͯ͠લड़ͷ k-means ʹΑΔ෼ੳΛߦͬͯਖ਼ղ཰. ͹Εͨɽඃ‫ ऀݧ‬B ʹؔͯ͠͸එͷ‫ے‬೑ʹՃ͑ͯൽෘిҐɼ. Λࢉग़ͨ͠ɽ. ಏ޸‫͕ܘ‬༗ޮಛ௃ྔͱͯ͠‫͛ڍ‬ΒΕͨɽൽෘిҐʹ͍ͭͯ ͸ɼԾઆͲ͓ΓʹΪΞൺେʹ͓͍ͯਫ਼ਆෛՙ͔Βྫྷ΍‫׼‬Λ. 4.2 ಛ௃ྔநग़݁Ռ. ͔͖ɼΨϧόχοΫൽෘ൓Ԡ͕༠ൃ͞ΕͨՄೳੑ͕͋Δɽ. ಛ௃ϕΫτϧͷ࣍‫ ਺ݩ‬m(1 < m < 6) ʹରͯ͠ඃ‫ ऀݧ‬3. ඃ‫ʹऀݧ‬Αͬͯ͸ൽෘిҐ͕ΪΞൺಛੑ൑ผʹ༗ޮͳಛ௃. ਓʹ‫ڞ‬௨ͷಛ௃ྔΛ༻͍ͨ৔߹ͷਖ਼ղ཰Λ Fig. 6 ʹࣔ͢ɽ. ྔͰ͋ΔՄೳੑ͕͋Δɽඃ‫ ऀݧ‬C ʹؔͯ͠͸ɼಛ௃ྔΛ. ॎ࣠͸ඃ‫ ऀݧ‬3 ਓͰ‫ڞ‬௨ͷಛ௃ྔΛ༻͍ͨ৔߹ͷਖ਼ղ཰ͷ. 4 ͭ࢖༻ͨ͠ࡍͷਖ਼ղ཰͸ 0.85 Ͱ͋Γɼಛ௃ྔΛ 5 ͭ࢖. ฏ‫ۉ‬஋Λ͍ࣔͯ͠Δɽॎ࣠͸ඃ‫ ऀݧ‬3 ਓͰ‫ڞ‬௨ͷಛ௃ྔ. ༻ͨ͠ࡍͷਖ਼ղ཰͸ 0.90 Ͱ͋ͬͨɽಛ௃ྔ 5 ͭͷ΄͏͕. c 2017 Information Processing Society of Japan . 4. 8.

(5) Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ද 2: ༗ޮಛ௃ྔͱ࢖༻ճ਺ Table 2 Usage count of effective features Part. Statistics. Count. Part. Statistics. Count. Corrugator supercilii. AveɾFreq. 5. Brachioradialis muscle. SDɾFreq. 2. Zygomaticus major. AveɾFreq. 2. blink count. -. 1. Masseter. -. 0. 0.3. 0.2 Normal High Ratio. Normal High Ratio. 0.18. 0.25. 0.16 0.14. 0.2. Force. Force. 0.12 0.15. 0.1 0.08. 0.1. 0.06 0.04. 0.05. 0.02 0. 0. 5. 10. 15. 20. 25. 30. 35. 0. 40. 0. 5. 10. 15. time [s]. 20. 25. 30. 35. time [s]. (a) Corrugator supercilii. (b) Zygomaticus major 0.25. 0.028 Normal High Ratio. 0.026. Normal High Ratio. 0.2. 0.024. 0.15. 0.02. Force. Force. 0.022. 0.018. 0.1. 0.016 0.014. 0.05. 0.012 0.01. 0. 5. 10. 15. 20. 25. 30. 35. 40. 0. 0. 5. 10. 15. 20. 25. time [s]. time [s]. (c) Masseter. (d) Brachioradialis muscle. 30. 35. ਤ 7: ‫ిے‬Ґͷৼ෯೾‫ܗ‬ Fig. 7 Myoelectric potential amplitude. ਖ਼ղ཰͕େ͖͘ͳͬͨɽ·ͣɼಛ௃ྔ 4 ͭͷ৔߹΋ 5 ͭͷ ৔߹΋ɼ͋͝ͷ‫ے‬೑ͷप೾਺੒෼ͱ๹ͷ‫ے‬೑ͷৼ෯ฏ‫ۉ‬஋ ͕༗ޮͰ͋ͬͨɽ͋͝ͷ‫ے‬೑͸ɼฏ‫ۉ‬஋͸༗ޮͰ͸ͳ͔ͬ ͕ͨɼർ࿑ͱؔ܎͕͋Δͱࢦఠ͞Ε͍ͯΔप೾਺੒෼ΛΈ Δ͜ͱͰ༗ޮಛ௃ྔͱͳΓಘΔ͜ͱΛࣔࠦͨ͠ɽඃ‫ ऀݧ‬C ʹ͍ͭͯ͸‫ిے‬ҐΛ‫ܭ‬ଌ͢ΔͷΈͰࣝผ͕ՄೳͰ͋Δ͜ͱ Λ͍ࣔࠦͯ͠Δɽ. 5. ͓ΘΓʹ ਤ 8: ඃ‫ऀݧ‬ຖʹ༗ޮಛ௃ྔΛద༻ͨ͠৔߹ͷਖ਼ղ཰ Fig. 8 Accuracy rate by suitable features for each subject. ࣗಈंͷӡస‫֮ײ‬Λ٬‫؍‬తʹධՁ͢ΔͨΊɼ‫ిے‬ҐΛ͸ ͡Ίͱ͢Δੜମ৴߸͔ΒΪΞൺಛੑͷมԽͱؔ࿈ͷ͋Δಛ ௃ྔΛநग़ͨ͠ɽද৘‫ͱے‬࿹ͷ‫ిے‬Ґɼࢹઢ৘ใɼൽෘ৘. c 2017 Information Processing Society of Japan . 5.

(6) Vol.2017-ITS-70 No.7 2017/8/4. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ใʹண໨͠ɼk-means ๏Λ༻͍ͯ༗ޮಛ௃ྔͷநग़Λߦͬ ͨɽͦͷ݁Ռɼᚧඑ‫ిے‬Ґͷৼ෯ฏ‫ͱۉ‬प೾਺੒෼͕ಛʹ ༗ޮͰ͋Δ͜ͱ͕ࣔ͞Εͨɽਖ਼ղ཰ͱͯ͠͸ɼඃ‫ऀݧ‬ผʹ. [16]. Jacobs KW, Hustmyer FE Jr, ʠEffects of Four Psychological Primary Colors on GSR, Heart Rate and Respiration Rate,ʡ Percept Mot Skills, Vol.38, No.3 pp763-766, 1974.. ಛ௃ྔΛબͼਪఆͨ͠৔߹ͷਖ਼ղ཰ͷඃ‫ऀݧ‬ฏ‫ ͕ۉ‬0.88ɼ શඃ‫ڞʹऀݧ‬௨ͨ͠ಛ௃ྔΛ༻͍ͯਪఆͨ͠৔߹ͷਖ਼ղ཰ ͕ 0.75 Ͱ͋ͬͨɽࠓ‫ޙ‬ͷ՝୊ͱͯ͠ɼଞͷγʔϯʹ͓͚ Δ‫ݕ‬౼ɼඃ‫਺ऀݧ‬ͷ૿Ճɼ࣮ं࣮‫ݧ‬ͷඞཁੑͳͲ͕‫͛ڍ‬Β ΕΔɽ ࢀߟจ‫ݙ‬ [1]. [2]. [3]. [4]. [5] [6]. [7]. [8]. [9]. [10]. [11]. [12]. [13]. [14]. [15]. Yoshihiko Kozawa, Gunji Sugimoto, Yasuhiko Suzuki, ʠA New Ride Comfort Meter,” SAE Technical Paper 86043, 1986. M. J. Griffin, E. M. Whitham, K. C. Parsons,ʠVibration and Comfort I. Translation Seat Vibration,ʡErgonomics, Vol.25, 1982, pp.603-630. ෢ҪҰ߶, ੴࠇ ཮༤, ʠ৐һͷ‫׭‬ೳධՁʹ΋ͱͮ͘৐Γ৺ ஍ධՁʡ, ๛ాதԝ‫ ॴڀݚ‬R&D ϨϏϡʔ, Vol.30, No.3, 1995, pp.47-56. J. Healey, R. Picard, ʠSmartCar: detecting driver stress,ʡ In Proceeding of IEEE 15th International Conference 2002, pp. 218-221. Ԭຊ༟࢘, ʠ‫ిے‬ҐଌఆʹΑΔࣗಈंͷ৐Γ৺஍ධՁʡ, ੜ࢈‫ڀݚ‬, Vol.62, No.3, 2010, pp.267-270. ૔৿ষ, ߴ‫ولޱ‬, ্ᑍਖ਼ٛ, ࠤ౉ࢁѥฌ, ਗ਼ਫٛ༤, ʠυ ϥΠόͷྗΈʹண໨ͨࣗ͠ಈंͷӡస͠΍͢͞ධՁ๏ʡ, ‫ײ‬ੑ޻ֶ‫ڀݚ‬࿦จू, Vol.6, No.2, pp.87-92, 2006. தଜ߂‫ؽ‬, த໺ެ඙, ํ๕, జਔ੒, େງਅ‫ܟ‬, ʠӡసऀ‫ۓ‬ ு౓ͱεςΞϦϯάάϦοϓྗͷ૬ؔʹؔ͢Δߟ࡯ʡ, ੜ ࢈‫ڀݚ‬, Vol.64, No.2, pp.269-272, 2012. Petty RE, Losch ME, Kim HS, ʠElectromyographic activity over facial muscle regions can differentiate the valence and intensity of affective reactions,ʡ Journal of Personality and Social Psychology Vol.50, No.2, pp.260268, 1986. Takayuki Sakurai, Masashi Toda, Shigeru Sakurazawa, Yuichi Nakamura,ʠDetection of Muscle Fatigue by the Surface Electromyogram and Its Application,ʡ IEEE International Conference on Computer and Information Science, pp.43-47, 2010. Janice Bagley, Leon Manelis, ʠEffect of Awareness on an Indicator of Cognitive Load,ʡ Perceptual and Motor Skills, Vol.49 (2), 1979. ࡾ୐ ৾࢘, ʠ঎඼։ൃɾධՁͷͨΊͷੜཧ‫ܭ‬ଌͱσʔλ ղੳϊ΢ϋ΢ʕੜཧࢦඪͷಛ௃ɺଌΓํɺ࣮‫ܭݧ‬ըɺσʔ λͷղऍɾධՁํ๏ʡ, NTS, 2017. Deubel H, Schneider WX, ʠSaccade Target Selection and Object Recognition: Evidence for a Common Attentional Mechanism,ʡ Vision Res, Vol.36, No.12 pp18271837, 1996. Bradley MM, Miccoli L, Escrig MA, Lang PJ, ʠThe pupil as a measure of emotional arousal and autonomic activation,ʡ Psychophysiology, Vol45, No.4, pp.602-607, 2008. দ໦༟ೋ, ࢤಊࣉ࿨ଇ, ২૲ཧ, ʠ։‫ܭ཰؟‬ଌΛ༻͍ͨӡ సऀͷ൓Ԡ࣌ؒͷ஗Ԇ༧ଌʹؔ͢Δ‫ڀݚ‬ʡ, ࣗಈंٕज़ձ ࿦จू, Vol.41, No.6, pp.1445-1450, 2010. H.Kataoka, H.Kano, H.Yoshida, A.Saijo, M.Yasuda, M.Osumi, ʠDevelopment of a skin temperature measuring system for non-contact stress evaluation,ʡ Engineering in Medicine and Biology Society, 1998.. c 2017 Information Processing Society of Japan . 6.

(7)