日本語音声におけるパワースペクトル因子の音声知覚上の役割

(1)

九州大学学術情報リポジトリ

Kyushu University Institutional Repository

日本語音声におけるパワースペクトル因子の音声知覚上の役割

岸田, 拓也

https://doi.org/10.15017/1931919

出版情報：Kyushu University, 2017, 博士（芸術工学）, 課程博士バージョン：

権利関係：

(2)

೔ຊޠԻ੠ʹ͓͚ΔύϫʔεϖΫτϧҼࢠͷԻ੠஌্֮ͷ໾ׂ

Perceptual roles of spectral-change factors in Japanese speech

؛ాɹ୓໵

Takuya Kishida

2018 ೥ 3 ݄

(3)

໨ ࣍

ୈ1ষ ং࿦ 1

1.1 ݚڀഎܠ . . . . 2

1.1.1 Իૉͷ஌֮ . . . . 3

1.1.2 ஌֮ͷख͕͔Γͷ৑௕ੑ . . . . 6

1.1.3 εϖΫτϧͷશମߏ଄͕΋ͭख͕͔Γ . . . . 11

1.1.4 ݴޠͷϦζϜͱ໐Իੑ . . . . 12

1.2 ຊ࿦จͷ໨త . . . . 15

1.3 ຊ࿦จͷߏ੒ . . . . 15

ୈ2ষ จԻ੠ͷ໌ྎͳ஌֮ʹཁ͢ΔύϫʔεϖΫτϧҼࢠͷݸ਺ 17 2.1 ୈ̎ষͷ໨త . . . . 17

2.2 ෼ੳ̍ɿ ى఺Ҡಈओ੒෼෼ੳʹΑΔύϫʔεϖΫτϧҼࢠͷநग़ . . . . 18

2.2.1 ෼ੳࢼྉ . . . . 18

2.2.2 खଓ͖ . . . . 19

2.2.3 ݁Ռͱߟ࡯ . . . . 25

2.3 ࣮ݧ̍ɿ ύϫʔεϖΫτϧҼࢠͷݸ਺ͱจԻ੠ͷ໌ྎ౓ͷؔ܎ . . . . 33

2.3.1 ࣮ݧࢀՃऀ . . . . 33

2.3.2 ࣮ݧ૷ஔ . . . . 33

2.3.3 ܹࢗԻ . . . . 34

2.3.4 खଓ͖ . . . . 41

2.3.5 ݁Ռͱߟ࡯ . . . . 42

ୈ3ষ ύϫʔεϖΫτϧҼࢠͷඇෛ௚ަجఈԽ 45 3.1 ୈ̏ষͷ໨త . . . . 45

3.2 ෼ੳ̎ɿ ύϫʔεϖΫτϧҼࢠͷඇෛ஋Խʹࡍ͢ΔӨڹͷྔతͳݕ౼ . . . . . 46

3.2.1 ඇෛ௚ަجఈԽͷํ๏ . . . . 46

3.2.2 ඇෛ௚ަجఈԽʹΑΔྦྷੵد༩཰ͷมԽ . . . . 46

(4)

3.3 ࣮ݧ̎ɿ ඇෛ௚ަجఈҼࢠΛ༻͍ͨจԻ੠໌ྎ౓ͷଌఆ . . . . 48

3.3.1 ࣮ݧࢀՃऀ . . . . 48

3.3.2 ࣮ݧ૷ஔ . . . . 48

3.3.3 ܹࢗԻ . . . . 49

3.3.4 खଓ͖ . . . . 49

3.3.5 ݁Ռͱߟ࡯ . . . . 49

ୈ4ষ ύϫʔεϖΫτϧҼࢠͷݸʑͷ໾ׂ 54 4.1 ୈ̐ষͷ໨త . . . . 54

4.2 ࣮ݧ̏ɿ ̐Ҽࢠ͔ΒͳΔύϫʔεϖΫτϧҼࢠͷݸʑͷ໾ׂ . . . . 55

4.2.1 ࣮ݧࢀՃऀ . . . . 55

4.2.2 ࣮ݧ૷ஔ . . . . 55

4.2.3 ܹࢗԻ . . . . 56

4.2.4 खଓ͖ . . . . 59

4.2.5 ݁Ռͱߟ࡯ . . . . 60

4.3 ࣮ݧ̐ɿ ̎Ҽࢠɺ̏Ҽࢠɺ̐Ҽࢠ͔ΒͳΔύϫʔεϖΫτϧҼࢠͷݸʑͷ໾ׂ 62 4.3.1 ࣮ݧࢀՃऀ . . . . 64

4.3.2 ࣮ݧ૷ஔ . . . . 64

4.3.3 ܹࢗԻ . . . . 64

4.3.4 खଓ͖ . . . . 68

4.3.5 ݁Ռͱߟ࡯ . . . . 68

ୈ5ষ ૯߹ߟ࡯ 72 5.1 ݁Ռͷུ֓ . . . . 72

5.2 ݁࿦ . . . . 73

5.3 ཧ࿦తҐஔ͚ͮ . . . . 76

5.3.1 ஌֮ͷख͕͔Γͷ৑௕ੑΛࣔͨ͠ઌߦݚڀͱͷؔ܎ . . . . 77

5.3.2 εϖΫτϧͷ౷ܭత෼ੳ͔ΒಘΒΕͨ΋ͷͷղऍ . . . . 78

5.3.3 Ի੠஌֮ͷཧ࿦ͱͷؔ܎ . . . . 78

5.3.4 ຊݚڀͷݶք . . . . 79

5.3.5 ຊݚڀͷԠ༻Մೳੑ . . . . 80

5.4 ࠓޙͷల๬ . . . . 80

(5)

จݙ 83

ँࣙ 88

෇ه 90

(6)

ୈ 1 ^ষ ^ং࿦

͸͡Ίʹ

ੜ෺Ͳ͏͠ʹΑΔίϛϡχέʔγϣϯ͸ࢹ֮ɺௌ֮ɺ৮֮ɺᄿ֮ͳͲɺ༷ʑͳײ֮Λ௨ͯ͠

ߦΘΕ͍ͯΔɻݴ༿Λ༻͍Δͱ͍͏ώτಛ༗ͷੑ࣭͔Βɺ༷ʑͳײ֮ͷதͰ΋ௌ֮Λ༻͍ͨί ϛϡχέʔγϣϯ͸զʑʹͱͬͯಛʹॏཁͰ͋Δͱݴ͑ΔͩΖ͏ɻݴ༿͸ओͱͯ͠Ի੠ʹΑͬ

ͯ఻͑ΒΕΔɻԻ͸ৼಈ͕ۭؾதΛ఻ൖ͢Δ෺ཧతͳݱ৅ʹա͗ͳ͍͕ɺώτ͸ௌ֮ث׭ͱൃ

੠ث׭ͱΛ޼Έʹར༻͢Δ͜ͱͰɺԻ੠Λ࢖ͬͯෳࡶͳίϛϡχέʔγϣϯ͕Ͱ͖ΔΑ͏ʹਐ Խͨ͠ɻ͜ͷෳࡶͳίϛϡχέʔγϣϯΛࢧ͑ΔԻ੠஌֮ͷ࢓૊ΈΛղ໌͢Δ͜ͱ͸ɺԻ੠ʹ

͔͔ΘΔݚڀ෼໺ʹ͓͚Δ࠷΋ॏཁͳςʔϚͷҰͭͰ͋Δͱݴ͑Δɻ

ຊ࿦จͰ͸ɺௌ֮ܥ຤ধʹ͓͚ΔԻ੠ͷεϖΫτϧදݱʹର͠౷ܭతख๏Λ༻͍Δ͜ͱͰͦ

ͷओཁͳ੒෼ΛऔΓग़͠ɺԻ੠஌֮ͷͨΊͷख͕͔Γͱͯͦ͠ΕΒͷ੒෼͕ͲͷΑ͏ʹར༻͞

Ε͍ͯΔͷ͔Λௌऔ࣮ݧʹΑͬͯௐ΂ͨɻԻ੠ͷԻڹతಛ௃ʹؚ·ΕΔԻ੠஌֮ͷख͕͔Γʹ

৑௕ੑ͕ͲΕ͚ͩ͋Δͷ͔Λɺ߹੒Ի੠Λ༻͍ͨௌऔ࣮ݧʹΑͬͯௐ΂Δݚڀͱɺ౷ܭతख๏

ʹΑΔ෼ੳΛ௨ͯ͠ௐ΂ΔݚڀͱΛ݁ͼ͚ͭΔݚڀͰ͋Δɻ

ຊষͰ͸ɺ·ͣɺԻ੠஌֮ʹؔ͢Δݚڀഎܠͱͯ͠ɺຊ࿦จͰऔΓѻͬͨ಺༰ͱಛʹؔ࿈ͷ

ਂ͍ઌߦݚڀΛ঺հ͢ΔɻઌߦݚڀʹΑͬͯ໌Β͔ͱͳͬͨ͜ͱΛ੔ཧ͢ΔதͰɺݕ౼͕ෆे

෼ͳ఺ΛऔΓ্͛Δɻͦͯ͠ɺຊ࿦จͷ໨తΛࣔ͢ɻ࠷ޙʹɺຊ࿦จશମͷߏ੒Λઆ໌͢Δɻ

(7)

1.1 ݚڀഎܠ

ɹ

ݴޠԻͱͯ͠ͷԻ੠Λ༻͍ͨௌ֮ίϛϡχέʔγϣϯͷ࿮૊Έʹ͸ɺ͍͔ͭ͘ͷஈ֊͕͋Δ ͱߟ͑ΒΕ͍ͯΔ(de Saussure, 1959)ɻਤ 1.1͸ɺͦͷௌ֮ίϛϡχέʔγϣϯͷ༷ʑͳஈ֊

Λ໛ࣜతʹࣔͨ͠΋ͷͰ͋Δɻ࿩ऀ͕ࣗ਎ͷߟ͑΍ײ৘Λଞऀʹ఻͑Α͏ͱ͢Δͱ͖ɺ·ͣ

͸ͦͷߟ͑΍ײ৘͕࿩ऀͷ೴಺ͰݴޠԽ͞ΕΔɻݴޠ͸࿩ऀͷ೴಺Ͱ͸ௌ֮Πϝʔδͱͯ͠

දݱ͞Ε͍ͯΔɻ͜ͷௌ֮ΠϝʔδΛ࣮ࡍͷ෺ཧతݱ৅ͱͯ͠ͷԻ೾ͱ࣮ͯ͠ݱ͢ΔͨΊʹɺ

೴͔ΒௐԻث׭΁ӡಈࢦྩ͕ૹΒΕɺൃ੠͕ى͖Δɻൃ੠ʹΑͬͯੜͨ͡Ի೾͕ௌऔऀͷࣖʹ

ೖྗ͞ΕΔͱɺۭؾͷৼಈͰ͋ͬͨԻ೾͕ௌ֮ܥͷॲཧΛܦͯௌऔऀʹௌ֮ΠϝʔδΛҾ͖ى

ͤ͜͞Δɻௌऔऀͷ೴಺Ͱ΋·ͨɺௌ֮Πϝʔδ͸ݴޠͦͷ΋ͷͱ݁ͼ͍͍ͭͯΔɻΑͬͯɺ

ௌऔऀʹ࿩ऀ͕఻͑Α͏ͱͨ͠ݴޠ಺༰͕఻ΘΔɻҎ্͕ௌ֮ίϛϡχέʔγϣϯͷجຊత

࿮૊ΈͰ͋ΔɻDenes and Pinson (1993)͸͜ͷ࿮૊Έͷதʹɺ࿩ऀ͕ൃ࿩ͨ͠Ի੠Λ࿩ऀࣗ

਎͕ௌऔ͢Δͱ͍͏ஈ֊΋͋Δ͜ͱΛड़΂͓ͯΓɺҰ࿈ͷஈ֊͕࠯ͷΑ͏ʹ࿈ͳΔ༷ࢠ͔Βɺ

ௌ֮ίϛϡχέʔγϣϯͷ࿮૊ΈΛݴ༿ͷ࠯ͱݺশ͍ͯ͠Δɻ͜ͷݴ༿ͷ࠯ʹ͓͍ͯɺԻ೾Λ

ௌ֮ΠϝʔδͱରԠ͚ͮΔஈ֊͕Ի੠஌֮Ͱ͋Δɻ

Linguistic level

Linguistic level Physiological

level

Acoustic level

Physiological level

Speaker Listener

Brain Vocal muscles Ear Brain

Ear

Sound wave

ਤ 1.1ௌ֮ίϛϡχέʔγϣϯͷ༷ʑͳஈ֊(ݴ༿ͷ࠯)ɻDenes and Pinson (1993)ʹܝࡌͷ ਤΛࢀߟʹචऀ͕࡞੒ɻ

Ի੠஌֮͸࣮ݧ৺ཧֶɾݴޠֶɾௌ֮ݚڀɾిؾ޻ֶɾਓ޻஌ೳݚڀͳͲͷ༷ʑͳ෼໺ͷݚ ڀऀ͕ؔ৺ΛدͤΔɺֶࡍతͳྖҬͰ͋Δ(Pisoni, 1985)ɻ༷ʑͳཱ৔ͷݚڀ͔ΒಘΒΕͨ஌

ݟ͸ɺ֎ࠃޠͷशಘɺޮ཰తͳԻ੠৴߸ͷ఻ૹٕज़ɺԻ੠ڧௐٕज़ɺࣗಈԻ੠ೝٕࣝज़ɺ೉ௌ

(8)

ऀͷ૷༻͢Δิௌثɾਓ޻಺ࣖͷ։ൃͳͲɺ๛͔ͳ฻Β͠Λઃܭ͢ΔͨΊʹར༻Ͱ͖Δɻ Ի੠ʹ͸ɺݴޠ಺༰Ҏ֎ʹ΋࿩ऀͷੑ࣭ɾঢ়ଶɾײ৘ͱ͍ͬͨ΋ͷΛ఻͑Δ໾ׂ͕͋Δ(Schuller

et al., 2013)͕ɺݴޠ಺༰ͷ఻ୡʹݶͬͯݴ͑͹ɺԻ੠஌֮ͷݚڀ͸͓͓Αͦ̎ͭͷݚڀྖҬ

ʹ෼͚ΒΕΔͩΖ͏ (Plomp, 2002; Samuel, 2011)ɻ̍ͭ໨͸ɺԻ੠ʹ͓͚Δ͋ΔԻڹతಛ௃

͕ɺԻૉɺԻઅɺ୯ޠͱ͍ͬͨԻ੠ͷߏ੒୯Ґͷ஌֮ͱͲͷΑ͏ʹؔ܎͍ͯ͠Δͷ͔Λௐ΂Δ

ྖҬͰ͋Δɻͦͯ͠΋͏̍ͭ͸ɺ೔ৗձ࿩ͷΑ͏ʹ࿈ଓతʹൃ࿩͞ΕͨԻ੠Λௌ͍ͯɺͲͷΑ

͏ʹͯͦ͠ͷԻ੠Λݴޠͱͯ͠ೝࣝ͠ɺॲཧ͍ͯ͠Δͷ͔Λௐ΂ΔྖҬͰ͋ΔɻຊݚڀͰ͸ɺ จ୯ҐͷԻ੠Λର৅ͱͯ͠ɺԻ੠৴߸ͷ෼ੳͱௌऔ࣮ݧΛߦͬͨɻ࿈ଓతʹൃ࿩͞ΕͨԻ੠Λ ѻ͏ͱ͍͏఺Ͱ͸ຊݚڀ͸ޙऀͷݚڀྖҬͰ͋Δͱ΋ݴ͑Δ͕ɺԻ੠ͷԻڹతಛ௃ʹ͍ͭͯ஌

֮ͱରԠ͚ͮΔͱ͍͏఺Ͱ͸લऀͷྖҬͱ΋ؔ࿈͕ਂ͍ɻͦ͜Ͱݚڀഎܠͱͯ͠ɺલऀͱޙऀ

ͷݚڀྖҬͷҧ͍ʹͩ͜ΘΒͣɺຊݚڀͱؔ࿈͕ਂ͍ઌߦݚڀΛ঺հ͢Δɻ

1.1.1 Իૉͷ஌֮

Իૉ͸Ի੠ͷ࠷খߏ੒୯ҐͰ͋ΔɻԻૉ͸฼ԻͱࢠԻͱʹେผͰ͖ɺڞ௨͢Δ෦෼͸͋Δ΋

ͷͷݴޠ͝ͱʹͦͷछྨͱମܥ͸গͣͭ͠ҟͳ͍ͬͯΔɻԻ੠஌֮ͷ࠷ॳظͷݚڀͰ͸ɺԻૉ

Λಉఆ͢ΔͷʹͲͷΑ͏ͳԻڹతಛ௃͕ඞཁͰ͋Δͷ͔͕ৄ͘͠ௐ΂ΒΕͨɻԻΛࢹ֮తʹͱ Β͑Δ͜ͱΛՄೳͱ͢Δɺα΢ϯυεϖΫτϩάϥϑ(Potter, 1945)ͱ͍͏૷ஔ͕͋Δɻ͜ͷ

૷ஔʹԻ੠৴߸Λೖྗ͢Δͱɼ৴߸͕ղੳ͞Εɺؚ·Ε͍ͯΔप೾਺੒෼ͷ࣌ؒมಈ͕ࢴ্ʹ ೱ୶Ͱඳ͔Εͨ΋ͷ(εϖΫτϩάϥϜ)Λग़ྗ͢Δɻਤ1.2͸Ի੠ͷ࣌ؒ೾ܗͱͦͷ࣌ؒ೾ܗ Λղੳ͢Δ͜ͱͰಘΒΕΔεϖΫτϩάϥϜͷྫͰ͋ΔɻҰํύλʔϯϓϨΠόοΫ(Cooper, Liberman, & Borst, 1951)͸ɺα΢ϯυεϖΫτϩάϥϑͱ͸ٯͷൃ૝ʹΑΔ΋ͷͰɺεϖΫ τϩάϥϜΛ໛ٖͯ͠ඳ͔ΕͨύλʔϯΛಡΈࠐΈɺԻΛ࠶ੜ͢Δͱ͍͏૷ஔͰ͋Δɻύλʔ ϯϓϨΠόοΫͰԻ੠Λ໛ٖͨ͠Իڹ৴߸͕࡞ΒΕɺ͜ΕΛ࢖ͬͨԻૉͷ஌࣮֮ݧ͕ߦΘΕ

ͨɻϋεΩϯεݚڀॴ (Haskins Laboratories)ͷݚڀऀʹΑͬͯߦΘΕͨ͜ΕΒ࠷ॳظͷݚڀ

͔ΒɺԻૉͷ஌֮ʹ͸ϑΥϧϚϯτ͕ॏཁͰ͋Δ͜ͱ͕໌Β͔ʹ͞Εͨ(Delattre, Liberman, Cooper, & Gerstman, 1952; Liberman, 1957)ɻ੠ଳͷৼಈͰੜͨ͡Ի೾͸ޱ৶͔Β์ࣹ͞Ε Δ·Ͱͷؒʹप೾਺͝ͱʹৼ෯͕ڧΊΒΕͨΓऑΊΒΕͨΓ͞ΕΔɻͲͷप೾਺ʹ͓͍ͯͲ ͷఔ౓ৼ෯ͷڧऑͷมԽ͕ى͖Δͷ͔͸ɺ੠ଳ͔Βޱ৶·Ͱͷܗঢ়ɺ͢ͳΘͪ੠ಓܗঢ়ʹΑΔ ڞৼಛੑͰܾ·ΔɻϑΥϧϚϯτ͸ͦͷΑ͏ʹͯ͠Ͱ͖ΔԻ੠ͷεϖΫτϧแབྷ্ͷࢁͷ͜ͱ Ͱ͋Δɻ ͦͷϐʔΫͱͳΔप೾਺͕ϑΥϧϚϯτप೾਺Ͱ͋Γɺ஋͕௿͍ॱʹୈ̍ϑΥϧϚ

(9)

ϯτɺୈ̎ϑΥϧϚϯτɺୈ̏ϑΥϧϚϯτͱ͍͏Α͏ʹݺ͹ΕΔɻ฼Իͷൃ੠࣌͸ϑΥϧϚ ϯτ͕ఆৗతʹ؍࡯͞ΕΔͨΊɺϑΥϧϚϯτप೾਺ͷ෼෍ͷύλʔϯΛௌ͖෼͚Δ͜ͱͰ

฼ԻΛ஌֮͢Δ͜ͱ͕Ͱ͖ɺҰํࢠԻ͸ϑΥϧϚϯτप೾਺ͷભҠύλʔϯ͕ಛ௃తͰ͋Γɺ

͜ΕΛख͕͔Γʹ஌֮Ͱ͖Δͱߟ͑ΒΕ͍ͯΔɻਤ1.3ʹ೔ຊޠͷ฼Իɺ/a/ͱ/i/ͷϑΥϧϚ ϯτͷύλʔϯͷҧ͍ΛݟΔ͜ͱ͕Ͱ͖Δɻ฼Իɺ/a/ͱ/i/ΛͦΕͧΕൃ੠͢Δࡍ͸੠ಓͷڱ ΊΒΕΔҐஔ͕ҟͳΔͨΊɺڞৼಛੑ͕มԽ͠ɺͦΕ͕ϑΥϧϚϯτύλʔϯ(εϖΫτϧแ བྷ)ͱͯ͠ݱΕΔɻҰํεϖΫτϧͷࡉ͔ͳมಈύλʔϯ(ඍࡉߏ଄)͸ɺ੠ଳৼಈʹΑͬͯ࡞

ΒΕΔͨΊɺಉ͡੠ͷߴ͞Ͱൃ੠͠Α͏ͱ͢Ε͹ɺͦͷมಈͷִؒ͸ೋͭͷ฼ԻͷؒͰ΋΄ͱ ΜͲมΘΒͳ͍ɻ

Time (s)

0 2.5

0

Time (s)

0 2.5

0 8000

Frequency (Hz)

0.5 1 1.5 2

1600 3200 4800 6400

Amplitude

ਤ 1.2 Ի੠ͷ࣌ؒ೾ܗ(্)͓ΑͼͦͷεϖΫτϩάϥϜ(Լ)ͷྫɻʮলΤωϧΪʔ͕ڣ͹Ε

͍ͯ·͢ɻʯͱ͍͏จ಺༰Λஉੑ͕ൃ࿩ͨ͠΋ͷɻNTT-ATଟݴޠԻ੠σʔλϕʔε2002ʹ ऩ࿥ɻ

(10)

0 2 4 6 8 10 12 14

0 1000 2000 3000 4000 5000 6000 7000 8000

Frequency (Hz)

Amplitude (dB)

/a/

/i/

ਤ 1.3೔ຊޠͷ฼ԻͷεϖΫτϧ(ࡉ͍࣮ઢ͕/a/ɺࡉ͍ഁઢ͕/i/)ͱͦͷεϖΫτϧแབྷ(ଠ

͍࣮ઢ͕/a/ɺଠ͍ഁઢ͕/i/)ɻචऀ͕ൃ੠ͨ͠΋ͷΛ࿥Իɺ෼ੳͨ͠ɻ

Իૉͷಉఆʹ͸ϑΥϧϚϯτ͕ॏཁͰ͋Δ͜ͱ͕෼͔͕ͬͨɺޙଓ͢Δ฼Իͷछྨ͕มΘ Δͱɺಉ͡ࢠԻͰ͋ͬͯ΋ͦͷϑΥϧϚϯτप೾਺ͷ࣌ؒભҠύλʔϯ͕ඇৗʹҟͳΔ͜ͱ

͕ಉ࣌ʹ؍࡯͞ΕɺಛఆͷࢠԻͰ͋ΔͱಉఆͰ͖ΔෆมతͳԻڹతಛ௃Λݟग़͢͜ͱ͸ࠔ೉

Ͱ͋Δ͜ͱ͕໌Β͔ͱͳͬͨɻͦ͜ͰɺLibermanΒ͸ɺԻૉͱௐԻӡಈͱͷؒͷҰ؏ͨ͠ର Ԡؔ܎ʹண໨ͯ͠ɺԻڹతಛ௃Ͱ͸ͳ͘ɺௐԻث׭Λಈ͔͢ے೑΁ͷӡಈࢦྩʹࢠԻΛಉఆ

͢Δෆมతಛ௃Λݟग़͢͜ͱ͕Ͱ͖ΔͩΖ͏ͱओுͨ͠(Liberman, Cooper, Shankweiler, &

Studdert-Kennedy, 1967)ɻ͜Ε͕Ի੠஌֮ͷӡಈཧ࿦(Motor Theory)Ͱ͋Δɻ͞Βʹӡಈཧ

࿦Ͱ͸ɺώτ͕Ի੠Λ஌֮͢Δͱ͖͸ͦΕҎ֎ͷԻͷ஌֮ͱ͸ผʹɺઐ༻ͷػߏΛར༻͍ͯ͠

Δͱओு͍ͯ͠Δɻӡಈཧ࿦Λࢧ࣋͢Δݚڀऀ͸ɺྫ͑͹ɺೋॏ஌֮(Rand, 1974)΍Χςΰ Ϧʔ஌֮(Liberman, Harris, Hoﬀman, & Griﬃth, 1957)ͱ͍ͬͨݱ৅Λൃݟ͠ɺ͜Ε͕Ի੠

ܹࢗΛ༻͍ͨͱ͖ʹ͚ͩಛ༗ʹى͜Δݱ৅Ͱ͋Δͱͯ͠ӡಈཧ࿦Λূ໌͠Α͏ͱͨ͠ɻԻ੠஌

֮ͷӡಈཧ࿦͸ɺޙͷԻ੠஌֮ݚڀʹ༩͑ͨӨڹͷେ͖͞Ώ͑ʹɺ࠷΋ॏཁͳཧ࿦ͷҰͭͰ͋

Δͱݴ͑Δɻ

ӡಈཧ࿦ʹର߅͢ΔܗͰɺԻ੠஌֮ͷཧ࿦͕ෳ਺ఏএ͞Ε͍ͯΔɻ͜͜Ͱ͸ɺBlumsteinͱ

Stevens͕ओு͢ΔɺԻڹతෆมੑཧ࿦ʹ͍ͭͯ৮ΕΔɻ͜ͷཧ࿦͸εϖΫτϧͷશମߏ଄͕

ख͕͔ΓͱͳΔͱ͍͏఺Ͱຊ࿦จͷ಺༰ͱؔ࿈͢Δɻӡಈཧ࿦ͷࢧ࣋ऀΒ͸ɺԻ੠ͷԻڹత ಛ௃ͷதʹԻૉΛಉఆͰ͖Δෆมతͳಛ௃Λݟग़͢͜ͱΛఘΊ͕ͨɺBlumstein and Stevens

(1979, 1980)͸ԻૉΛೋ߲ରཱૉੑΛ࢖ͬͯ෼ྨ͢Δͱ͍͏ߟ͑ํʹैͬͯɺԻ੠ͷԻڹత

(11)

ಛ௃͔ΒԻૉΛಉఆ͢Δෆมతಛ௃Λݟ͚ͭΑ͏ͱͨ͠ɻྫ͑͹ด࠯ࢠԻͷด࠯ͷ։์͔Β

20–30 ms·Ͱͷ۠ؒͷεϖΫτϧʹ͓͍ͯɺΤωϧΪʔ͕प೾਺্࣠ͷத৺ʹू໿͍ͯ͠Δ

͔શମతʹ֦ࢄ͍ͯ͠Δ͔(ू໿ੑ–֦ࢄੑͷରཱ)ɺ֦ͦͯ͠ࢄ͍ͯ͠Δ৔߹͸ɺߴҬʹ޲͏

ʹͭΕͯϑΥϧϚϯτͷϐʔΫͷৼ෯͕૿Ճ͍ͯ͠Δ͔ݮগ͍ͯ͠Δ͔(ߴԻௐੑ–௿Իௐੑ

ͷରཱ)Λ؍࡯͢Δ͜ͱʹΑͬͯɺด࠯ࢠԻͷௐԻҐஔʹΑΔ̏छྨͷ෼ྨ͕ՄೳͰ͋Δ͜ͱ Λࣔͨ͠ɻ͔͠͠ҰํͰɺෆมతͳಛ௃ͱಉ࣌ʹޙଓ฼Իͷҧ͍ʹΑͬͯมΘΔಛ௃͕༩͑Β Εͨͱ͖͸ɺௌऔऀ͸มΘΔಛ௃ͷํΛجʹͯ͠ԻૉΛ൑அ͢Δ܏޲͕͋Δ͜ͱ͕ࣔ͞Ε͓ͯ

Γ(Blumstein, Isaacs, & Mertus, 1982; Walley & Carrell, 1983)ɺԻڹతෆมੑཧ࿦͸׬શͱ

͍͏Θ͚Ͱ͸ͳ͍ɻ

ଞʹ΋Ի੠஌֮ͷཧ࿦ʹ͸୅දతͳ΋ͷͱͯ͠ɺ௚઀࣮ࡏཧ࿦ (Direct Realism Theory;

Fowler, 1991), ҰൠΞϓϩʔν(General Approach; Diehl, Lotto, & Holt, 2004)ͳͲ͕͋Δɻ

͜ΕΒͷཧ࿦͸ɺೋॏ஌֮ݱ৅͕ඇԻ੠Ͱ΋ੜ͡Δ͜ͱ(Fowler & Rosenblum, 1990)΍ɺΧ ςΰϦʔ஌͕֮ώτҎ֎ͷੜ෺ʹ΋ΈΒΕΔ(Kluender, Diehl, Killeen, et al., 1987)ͱ͍͏࣮

ݧใࠂΛجʹཱͯΒΕͨ΋ͷͰ͋ΔɻԻ੠஌֮ͷཧ࿦ʹ͓͚ΔͦΕͧΕͷߟ͑ํ͸ɺ஌֮ͷର

৅͕ௐԻӡಈͰ͋Δ͔Ͳ͏͔ɺԻ੠஌͕֮ઐ༻ͷػߏʹΑΔ΋ͷ͔Ͳ͏͔ͱ͍͏ཱ৔Ͱ෼ྨͰ

͖Δ(Diehl et al., 2004)ɻ͔͠͠ɺ୭͠΋͕ೲಘ͢Δཧ࿦͸ͳ͘ɺݱࡏͰ΋׆ൃͳٞ࿦͕ଓ͍

͍ͯΔɻ

1.1.2 ஌֮ͷख͕͔Γͷ৑௕ੑ

࣮؀ڥʹ͸Ի੠Ҏ֎ʹ΋༷ʑͳԻ͕͋;Ε͓ͯΓɺձ࿩தʹແؔ܎ͷԻΛԻ੠ͱಉ࣌ʹௌ͘

ঢ়گ͸͠͹͠͹͋Δɻ͞Βʹ͸ɺԻ͸͙͢ʹཱͪফ͑ͯ͠·͏΋ͷͰ͋ΔɻΑͬͯԻ੠Λ࢖ͬ

ͯ҆ఆͨ͠ίϛϡχέʔγϣϯΛߦ͏ͨΊʹ͸ɺԻ੠ʹ͋Δఔ౓༨৒ʹ஌֮ͷख͕͔ΓͱͳΔ

৘ใؚ͕·Ε͍͔ͯͯ͠Δ΂͖Ͱ͋Δɻ࣮ࡍʹɺԻ੠ʹ͓͚Δ஌֮ͷख͕͔Γͷ৑௕ੑΛࣔ͢

༷ʑͳใࠂ͕ͳ͞Ε͍ͯΔɻ͜͜Ͱ͸प೾਺৘ใʹؔ͢Δ৑௕ੑʹ͍ͭͯ঺հ͢Δɻ

಺ࣖͷ૊৫Ͱ͋Δ᥾ڇʹԻ͕ೖྗ͞Εͨͱ͖ɺҟͳΔप೾਺͝ͱʹ᥾ڇͷجఈບ্Ͱڧ͘

ৼಈ͢Δ৔ॴ͕ҟͳΔ͜ͱ͔Βɺզʑͷࣖ͸प೾਺෼ੳثͱͯ͠͸ͨΒ͘͜ͱ͕໌Β͔ʹ͞

Ε͍ͯΔ(Schnupp, Nelken, & King, 2011; Plack, 2014)ɻ͜ͷௌ֮ܥͷप೾਺෼ੳػೳ͸த

৺प೾਺ͱଳҬ෯ͷ͜ͱͳΔϑΟϧλ͕ෳ਺ฒΜͩϑΟϧλόϯΫͱͯ͠ϞσϧԽͰ͖Δɻ

͜ͷϑΟϧλ͕ௌ֮ϑΟϧλ(Patterson, 1974; Unoki, Irino, Glasberg, Moore, & Patterson, 2006; Moore, 2013)ͱݺ͹ΕΔ΋ͷͰ͋ΔɻྟքଳҬ(Fletcher, 1940; Zwicker & Terhardt,

(12)

1980; Greenwood, 1990; Schneider, Morrongiello, & Trehub, 1990)͸ௌ֮ϑΟϧλΛۣܗʹۙ

ࣅͨ͠΋ͷͰ͋Δɻಉ࣌ϚεΩϯάΛར༻ͨ͠ௌऔ࣮ݧ͔ΒɺྟքଳҬͷଳҬ෯͕ٻΊΒΕΔ (Fletcher, 1940; Zwicker & Terhardt, 1980)ɻྟքଳҬ෯͸த৺प೾਺͕໿500 Hz·Ͱ͸Ұఆ

ͷ100 Hzఔ౓Ͱ͋Δ͕ɺ500 HzҎ্Ͱ͸͓Αͦத৺प೾਺ͷ20%ͷ޿͞ʹͳΔͱ͍͏ಛ௃

Λ΋ͭɻ͔͠͠ௌ֮ϑΟϧλͷܗঢ়͸࣮ࡍʹ͸ۣܗͰ͸ͳ͍ɻPatterson (1974)ʹΑͬͯௌ֮

ϑΟϧλͷܗঢ়ΛٻΊΔ࣮ݧख๏͕ఏҊ͞Εɺௌ֮ϑΟϧλͷܗঢ়͕໌Β͔ʹͳ͍ͬͯͬͨɻ

ௌ֮ϑΟϧλ͸த৺प೾਺ʹରͯ͠ରশͳܗঢ়Ͱ͸ͳ͘ɺ௿Ҭଆ͸ͳͩΒ͔ʹϑΟϧλग़ྗ͕

௿Լ͠ɺߴҬଆ͸ٸफ़ʹϑΟϧλग़ྗ͕௿Լ͢Δͱ͍͏ಛ௃͕͋Δɻ͞Βʹɺ௿Ҭʹ͓͍ͯҰ ఆͷଳҬ෯ͱߟ͑ΒΕ͍ͯͨௌ֮ϑΟϧλ͸࣮ࡍʹ͸௿Ҭ΄Ͳڱ͘ͳΔͱ͍͏͜ͱ΋໌Β͔

ͱͳͬͨɻ·ͨɺௌ֮ϑΟϧλͷܗঢ়͸ೖྗϨϕϧʹΑͬͯ΋มԽ͢Δ͜ͱ΋෼͔͍ͬͯΔɻ தఔ౓ͷϨϕϧʹ͓͍ͯ͸ɺϑΟϧλग़ྗ͸ର਺प೾਺্࣠ͰରশͰ͋Δͱߟ͑ͯΑ͍ɻௌ֮

ϑΟϧλͷଳҬ෯Λ౳ՁۣܗଳҬ෯ʹ׵ࢉ͢Δ͜ͱ͸༗༻Ͱ͋Δɻ౳ՁۣܗଳҬ෯ͱ͸ɺௌ֮

ϑΟϧλ͕௨͢ന৭ࡶԻͷύϫʔͱಉྔͷύϫʔΛ௨͢Α͏ͳۣܗϑΟϧλͷଳҬ෯ͷ͜ͱ

Ͱ͋Δ(Moore, 2013)ɻۣܗϑΟϧλͷߴ͞͸ௌ֮ϑΟϧλͷத৺प೾਺ʹ͓͚Δߴ͞ʹͦΖ

͑ΒΕΔɻਤ1.4Ͱதఔ౓ͷϨϕϧͷԻʹର͢Δௌ֮ϑΟϧλͷ౳ՁۣܗଳҬ෯ͱྟքଳҬ෯ ͱΛൺֱ͢Δ͜ͱ͕Ͱ͖ΔɻൺֱతߴଳҬʹ͓͍ͯ͸ɺௌ֮ϑΟϧλͷ౳ՁۣܗଳҬ෯ͱྟք ଳҬ෯͸ಉఔ౓Ͱ͋Δ͜ͱ͕Θ͔Δɻ ྟքଳҬͷଳҬ෯͸جఈບͷ1.3 mm෼ͷ௕͞ʹରԠ

͠(Fastl & Zwicker, 2006)ɺ͜Ε͸ଳҬͷத৺प೾਺ͷ͓Αͦ1/4ʙ1/3ΦΫλʔϒఔ౓Ͱ͋

Δ(Plomp, 2002)ɻ͜Ε͕ௌ֮ͷप೾਺෼ղೳͰ͋Δ͕ɺ࣍ʹ͍͔ࣔͭ͘͢ͷྫͷΑ͏ʹɺԻ

੠ͷ஌֮ʹ͓͍ͯ͸͜ͷ෼ղೳ͸े෼͗͢Δɻ

(13)

10 100 1000

100 500 1000 5000 10000

50 50 500

Center frequnecy (Hz)

Bandwidth (Hz)

Critical bandwidth

Equivalent rectangular bandwidth

ਤ 1.4ྟքଳҬ෯ͱௌ֮ϑΟϧλͷ౳ՁۣܗଳҬ෯ͷൺֱɻ྆࣠ͱ΋ʹର਺࣠ɻZwicker and Terhardt (1980)͓ΑͼɺMoore (2013)Λࢀߟʹ࡞੒ͨ͠ɻ

Ի੠͸͓Αͦ50–8000 Hzͷप೾਺ଳҬʹΤωϧΪʔ͕෼෍͓ͯ͠ΓɺͦͷൣғʹԻ੠஌֮

ͷͨΊͷ༷ʑͳख͕͔Γ͕༩͑ΒΕ͍ͯΔɻಉ͡ःஅप೾਺ʹ͓͍ͯ௿Ҭ௨աϑΟϧλʹ௨

͞ΕͨԻ੠ͱߴҬ௨աϑΟϧλʹ௨͞ΕͨԻ੠ͱͷ̎ͭͷ৚݅ͰɺԻ੠ͷ໌ྎ౓͕ःஅप೾

਺ͷมԽͱͱ΋ʹͲͷΑ͏ʹมԽ͢Δ͔Λௐ΂ͨݚڀ͕ෳ਺͋Δ(French & Steinberg, 1947;

Hirsh, Reynolds, & Joseph, 1954; Miller & Nicely, 1955; Studebaker, Pavlovic, & Sherbecoe, 1987)ɻःஅप೾਺͕௿͍৔߹͸ɺߴҬ௨աϑΟϧλʹ௨͞ΕͨԻ੠ͷํ͕໌ྎ౓͕ߴ͘ɺͦ

ͷٯͷ৔߹͸௿Ҭ௨աϑΟϧλʹ௨͞ΕͨԻ੠ͷํ͕໌ྎ౓͕ߴ͘ͳΔΘ͚͕ͩɺ̎ͭͷ৚݅

Ͱಉ͡໌ྎ౓ͱͳΔͱ͜Ζͷःஅप೾਺(͓Αͦ1700 Hz෇ۙ)ʹ͓͍ͯɺଟ͘ͷݚڀͰԻઅ ਖ਼౴཰΍୯ޠਖ਼౴཰͕50%Ҏ্ͱͳΔ͜ͱ͕ใࠂ͞Ε͍ͯΔɻ͜ͷ͜ͱ͸૬ิతͳԻڹత৘

ใͦΕͧΕ͚ͩΛ༻͍ͯɺԻઅ͋Δ͍͸୯ޠͷ஌͕֮͋Δఔ౓ՄೳͰ͋Δ͜ͱΛ͓ࣔͯ͠Γɺ

஌֮ͷख͕͔Γͷ৑௕ੑΛࣔ͢ҰͭͷྫͰ͋Δɻ·ͨؔ࿈͢Δݚڀͱͯ͠ɺ1/3ΦΫλʔϒͷ

޿͞ͷڱଳҬϑΟϧλʹ௨͞ΕͨԻ੠৴߸Ͱ΋ɺඇৗʹߴ͍໌ྎ౓͕ಘΒΕΔ͜ͱ͕ɺ͞Β

(14)

ʹ1/20ΦΫλʔϒͷΑΓڱ͍ଳҬϑΟϧλʹ௨͞ΕͨԻ੠৴߸Ͱ͋ͬͯ΋ɺϑΟϧλͷத৺

प೾਺͕1500 Hz෇ۙͰ͋Ε͹૬౰ߴ͍໌ྎ౓͕ಘΒΕΔ͜ͱ͕ Warren, Riener, Bashford, and Brubaker (1995)ʹΑͬͯใࠂ͞Ε͍ͯΔɻ

্ʹ͋͛ͨप೾਺ଳҬΛ੍ݶ͢ΔݚڀͷଞʹɺεϖΫτϧશମʹؚ·ΕΔ৘ใΛྼԽͤͨ͞Ի

੠ʹΑΔௌऔ࣮ݧͰ΋ɺԻ੠஌֮ͷख͕͔Γʹ৑௕ੑ͕͋Δ͜ͱ͕ใࠂ͞Ε͍ͯΔɻTer Keurs,

Festen, and Plomp (1992, 1993)͸ɺԻ੠ͷεϖΫτϧแབྷͷมԽΛΨ΢εϑΟϧλΛ࢖ͬͯ

ಷΒͤɺεϖΫτϧแབྷ্ͷࢁ͕௿͘ɺ୩͕ઙ͘ͳͬͨԻ੠͕ఆৗࡶԻԼͰͲΕ͚ͩ໌ྎʹௌ

͖ͱΕΔ͔Λௐ΂ͨɻΨ΢εϑΟϧλͷଳҬ෯Λม͑ͯɺεϖΫτϧแབྷͷมԽͷಷ͞ͷҟͳ ΔԻ੠Ͱൺֱͨ͠ͱ͜ΖɺΨ΢εϑΟϧλͷଳҬ෯͕1/3ΦΫλʔϒ·ͰͷԻ੠͸ɺॲཧΛߦ Θͳ͍৚݅ͷԻ੠ͱಉ౳ʹ໌ྎͰ͋Δ͜ͱ͕෼͔ͬͨɻͦΕ͚ͩͰͳ͘ɺ̐ΦΫλʔϒͷଳҬ ෯ͰεϖΫτϧแབྷΛಷΒͤͯ΋ɺࡶԻͷϨϕϧʹରͯ͠े෼ʹԻ੠ͷϨϕϧΛେ͖͘͢Ε

͹ɺ໌ྎʹԻ੠Λ஌֮Ͱ͖Δ͜ͱ΋෼͔ͬͨɻ͜͜Ͱ൴Β͸ɺԻ੠͕உੑͷ΋ͷͰ͋ͬͯ΋ঁ

ੑͷ΋ͷͰ͋ͬͯ΋݁Ռ͕มΘΒͳ͍ͱ͍͏͜ͱ͔ΒɺεϖΫτϧʹؚ·ΕΔඍࡉߏ଄ΑΓ΋

εϖΫτϧશମͷแབྷߏ଄͕Ի੠ͷ໌ྎ౓Λܾఆ͚ͮΔཁҼͰ͋Δͱߟ࡯͍ͯ͠Δɻ

εϖΫτϧશମͷ৘ใΛྼԽͤͨ͞Ի੠ͷผͷྫͱͯ͠ɺνϟϯωϧϘίʔμԻ੠(Dudley,

1939)͕͋ΔɻνϟϯωϧϘίʔμԻ੠͸ɺԻ੠Λ͍͔ͭ͘ͷप೾਺ଳҬ(νϟϯωϧ)ʹ෼͚

ͯॲཧΛ͢Δ͜ͱͰɺͦͷνϟϯωϧʹ͓͚Δৼ෯แབྷͷΈΛऔΓग़͠ɺͦͷऔΓग़͞Εͨ

ৼ෯แབྷͰผͷ৴߸(ൖૹ৴߸)ͷରԠ͢ΔνϟϯωϧΛͦΕͧΕۦಈ͢Δ͜ͱͰ߹੒͞ΕΔ Ի੠৴߸Ͱ͋Δ(ਤ1.5)ɻ ൖૹ৴߸͕ଳҬࡶԻͷ৔߹͸ࡶԻۦಈԻ੠(noise-vocoded speech)ɺ ਖ਼ݭ೾ͷ৔߹͸ਖ਼ݭ೾ۦಈԻ੠(sine-vocoded speech)ͱݺ͹ΕΔɻνϟϯωϧϘίʔμԻ੠

͸νϟϯωϧ਺ΛมԽͤ͞Δ͜ͱͰஈ֊తʹεϖΫτϧ৘ใΛྼԽͤ͞Δ͜ͱ͕Ͱ͖Δɻνϟ ϯωϧϘίʔμԻ੠Λ༻͍ͨݚڀͷ੒Ռ͸ਓ޻಺ࣖͷप೾਺νϟϯωϧͷઃఆͳͲʹར༻͞Ε

͍ͯΔ(Xu & Pﬁngst, 2008)ɻ

(15)

Original speech

Carrier signal BPF 1

Σ

BPF 2

BPF N

BPF 1 BPF 2 BPF N

LPF

Vocoded speech Rect.

Rect.

…

… … …

…

・

・・

ਤ 1.5 Nνϟϯωϧ͔ΒͳΔνϟϯωϧϘίʔμԻ੠ͷ࡞੒खॱΛࣔ͢ྲྀΕਤɻத৺प೾਺

ͷҟͳΔNݸͷଳҬ௨աϑΟϧλ(ਤதͷBPF)ʹ௨͞ΕͨͦΕͧΕͷԻ੠৴߸Λɺ͞Βʹ੔

ྲྀ(ਤதͷRect.)͠ɺ௿Ҭ௨աϑΟϧλ(ਤதͷLPF)ʹ௨͢͜ͱͰ֤प೾਺ଳҬʹ͓͚ΔԻ

੠৴߸ͷৼ෯แབྷ͕ಘΒΕΔɻ͜ͷৼ෯แབྷͰൖૹ৴߸(ࡶԻ·ͨ͸ਖ਼ݭ೾౳)ͷରԠ͢Δप

೾਺ଳҬΛৼ෯มௐ͠ɺ֤ଳҬͷ৴߸Λ଍͠߹ΘͤΔ͜ͱͰνϟϯωϧϘίʔμԻ੠͕߹੒͞

ΕΔɻ

Shannon, Zeng, Kamath, Wygonski, and Ekelid (1995)ͷ࣮ݧͰ͸ɺԻ੠Λ4000 HzҎԼʹ ଳҬ੍ݶͨ͠͏͑Ͱɺ̐νϟϯωϧͷࡶԻۦಈԻ੠ͱͯ͠߹੒ͨ͠৔߹Ͱ΋ɺจԻ੠Ͱ͋Ε

͹୯ޠਖ਼౴཰͕90%Λ௒͑Δ͜ͱ͕ใࠂ͞Ε͍ͯΔɻ·ͨɺࡶԻۦಈԻ੠ͱਖ਼ݭ೾ۦಈԻ੠

ͱΛൺֱ࣮ͨ͠ݧͰ͸ɺͲͪΒͷ৚݅Ͱ΋̐νϟϯωϧͰจԻ੠ͷ୯ޠਖ਼౴཰͕90%Λ௒͑

Δ݁Ռ͕ಘΒΕ͍ͯΔ(Dorman, Loizou, & Rainey, 1997)ɻ͞Βʹಉ༷ͷ࣮ݧ͸ɺෳ਺ͷ࿩ऀ

Λ࢖͏(Loizou, Dorman, & Tu, 1999)ɺए೥ऀͱߴྸऀͷ̎ͭͷ࣮ݧࢀՃऀάϧʔϓʹ෼͚Δ (Sheldon, Pichora-Fuller, & Schneider, 2008)ɺνϟϯωϧ಺ͷ࣌ؒ৘ใΛมԽͤ͞Δ(Souza &

Rosen, 2009),ݴޠ΍νϟϯωϧͷःஅप೾਺Λม͑Δ(Ellermeier, Kattner, Ueda, Doumoto,

& Nakajima, 2015)ͳͲɺ༷ʑͳ৚݅ͰߦΘΕ͖͕ͯͨɺ͍ͣΕͷ࣮ݧʹ͓͍ͯ΋νϟϯωϧ

ϘίʔμԻ੠͸̐ʙ̒ଳҬఔ౓͋Ε͹े෼ʹ໌ྎʹͳΔ͜ͱ͕෼͔͍ͬͯΔɻ͜ͷΑ͏ʹগ ͳ͍ଳҬ਺ͷνϟϯωϧϘίʔμԻ੠Ͱ͜Ε͚ͩ໌ྎʹԻ੠Λ஌֮Ͱ͖Δͷ͸ɺνϟϯωϧ

಺ͷ࣌ؒมԽͷ৘ใ͕ୈ̍ͷ஌֮ͷख͕͔ΓͰ͋Δ͔Βͩͱߟ࡯͞Ε͍ͯΔ(Shannon et al.,

1995)͕ɺνϟϯωϧؒͷϨϕϧࠩʹΑͬͯ༩͑ΒΕΔεϖΫτϧͷେ·͔ͳߏ଄͕ख͕͔Γ

(16)

ͱͳ͍ͬͯΔͱใࠂ͢Δݚڀ΋͋Δ(Roberts, Summers, & Bailey, 2010)ɻνϟϯωϧϘίʔ μԻ੠Λ༻͍ͨݚڀ͸ɺͲΕ͚ͩԻ੠ʹ৑௕ੑ͕͋Δͷ͔Λ͖ࣔͯͨ͠ɻ͔͠͠ͳ͕Βɺͳͥ

νϟϯωϧϘίʔμԻ੠Ͱ໌ྎʹԻ੠Λ஌֮͢Δ͜ͱ͕Ͱ͖Δͷ͔ͱ͍͏໰͍ʹରͯ͠͸े෼

ʹ౴͑Δ͜ͱ͸Ͱ͖͍ͯͳ͍ɻ·ͨɺͲͷप೾਺ଳҬͷ৘ใͷد༩͕໌ྎ౓ʹ༩͑ΔޮՌ͕େ

͖͍͔ʹ͍ͭͯ΋े෼ʹݕ౼͞Ε͍ͯͳ͍ɻ

1.1.3 εϖΫτϧͷશମߏ଄͕΋ͭख͕͔Γ

͜͜·ͰɺԻ੠஌֮ʹ͓͍ͯ͸ɺεϖΫτϧͷશମߏ଄͕ख͕͔ΓͰ͋Δͱ͍͏ՄೳੑΛ܁

Γฦ͖ࣔͯͨ͠͠ɻ͜ͷΑ͏ͳεϖΫτϧͷશମߏ଄͕΋ͭख͕͔Γʹ͍ͭͯผͷ֯౓͔Β ௐ΂Δํ๏ͱͯ͠ɺԻ੠ͷԻڹతಛ௃Λ౷ܭతख๏ʹΑͬͯ෼ੳ͢Δݚڀ͕͋Δɻ͜͜Ͱ͸ɺ

౷ܭతख๏ʹΑΔԻ੠ͷ෼ੳʹ͍ͭͯͷઌߦݚڀΛ঺հ͠ɺຊ࿦จͷςʔϚͱͳΔύϫʔεϖ ΫτϧҼࢠʹ͍ͭͯಋೖ͢Δɻ

Plomp, Pols, and Geer (1967)͸Φϥϯμޠͷ15ͷ฼ԻͷεϖΫτϧΛྟքଳҬ෯ʹ͍ۙɺ 1/3ΦΫλʔϒόϯυͰ18ଳҬʹ෼ׂ͠ɺͦΕͧΕͷଳҬͷύϫʔΛجʹओ੒෼෼ੳΛߦͬͨɻ

൴Β͸ओ੒෼෼ੳʹΑͬͯɺεϖΫτϧͷશମతͳಛ௃͕ͲͷΑ͏ͳ୯७ͳύλʔϯʹΑͬͯ

ߏ੒͞ΕΔͷ͔Λ͔֬ΊͨͷͰ͋Δɻͦͷ݁Ռɺୈ2ओ੒෼·ͰͰσʔλͷ͓Αͦ70%͕આ

໌Ͱ͖Δͱ෼͔Γɺ15ͷ฼ԻΛୈ̍ɺୈ̎ओ੒෼ۭؒͰे෼ʹ۠ผՄೳͰ͋Δ͜ͱ͕ࣔ͞Ε

ͨɻ·ͨɺ͜ͷୈ̍ɺୈ̎ओ੒෼্ۭؒʹ͓͚Δ฼Իͷ഑ஔ͸ɺୈ̍ϑΥϧϚϯτप೾਺ͱୈ

̎ϑΥϧϚϯτप೾਺ͷର਺஋Λ࣠ͱ͢Δฏ໘্ʹ͓͚Δ฼Իͷ഑ஔͱରԠ෇͚ΒΕΔ͜ͱ͕

෼͔ͬͨ(Pols, Tromp, & Plomp, 1973; Plomp, 1976, 2002)ɻ͜ͷΑ͏ͳσʔλΛഎܠʹɺ฼

Իͷࣝผʹ͓͍ͯɺϑΥϧϚϯτप೾਺ΑΓ΋Ή͠ΖεϖΫτϧશମͷܗঢ়͕ख͕͔Γͱͯ͠

༗༻Ͱ͋Δ͜ͱΛ Zahorian and Jagharghi (1993)͕͍ࣔͯ͠Δɻ

Ueda and Nakajima (2017)͸ Plomp et al. (1967); Pols et al. (1973); Plomp (1976, 2002) ͷ෼ੳख๏Λɺ̔ͭͷҟͳΔݴޠɾํݴʹ͓͚Δ࿈ଓతʹൃ࿩͞ΕͨจԻ੠Λର৅ʹ֦ு͠

ͨɻ൴Β͸ɺZwicker and Terhardt (1980)Λࢀߟͱͯ͠ઃఆͨ͠20ͷྟքଳҬͰɺจԻ੠ͷ ύϫʔεϖΫτϧͷ࣌ؒมಈΛ෼ׂ͠ɺ1 msຖʹ20ଳҬͷύϫʔ஋ͱͯ͠औΓग़͞Εͨ΋ͷ ΛҼࢠ෼ੳʹ͔͚ͨɻ͜ͷ෼ੳʹΑͬͯ̏ͭͳ͍̐ͭ͠ͷҼࢠΛऔΓग़͢ͱɺ̔ͭશͯͷݴޠ ʹ͓͍ͯڞ௨͢ΔύλʔϯͷҼࢠ͕ಘΒΕΔ͜ͱΛ൴Β͸ൃݟͨ͠ɻ͜ͷ݁Ռ͸ݴޠΛ௒͑ͨ

ීวతͳԻڹతಛ௃͕Ի੠ʹؚ·Ε͍ͯΔ͜ͱΛࣔ͢΋ͷͰ͋Δɻຊ࿦จͰ͸͜ͷҼࢠΛɺύ ϫʔεϖΫτϧͷ࣌ؒมಈΛߏ੒͢Δ͜ͱ͔ΒɺʮύϫʔεϖΫτϧҼࢠʯͱݺͿ͜ͱͱ͢Δɻ

(17)

͜Ε·Ͱʹ͓͍ͯɺԻ੠ͷεϖΫτϧͷશମతͳܗঢ়ͷ౷ܭత෼ੳ͔ΒಘΒΕΔεϖΫτϧ Λߏ੒͢Δओཁͳಛ௃͕ɺԻ੠஌֮ͷख͕͔Γͱͯ͠ͲͷΑ͏ʹػೳ͢Δ͔ʹ͍ͭͯௐ΂ͨݚ ڀ͸චऀ͕୳͢ൣғͰ͸ݟ౰ͨΒͳ͍ɻ͜ͷΑ͏ͳ౷ܭతख๏Λ༻͍ͨԻ੠ͷεϖΫτϧͷ

෼ੳ݁ՌΛɺԻ੠஌֮ͷ࢓૊Έͱ௚઀݁ͼ͚ͭͯߟ࡯͢Δͱ͍͏ΑΓ΋ɺऔΓग़͞Εͨओ੒෼

ۭؒΛ༻͍ͯ฼ԻΛࣗಈతʹࣝผ͢Δٕज़ʹར༻͢Δํ޲ʹݚڀ͕ൃల͍ͯ͠ΔΑ͏Ͱ͋Δɻ

Ueda and Nakajima (2017)͕औΓग़ͨ̏ͭ͠·ͨ͸̐ͭͷύϫʔεϖΫτϧҼࢠ͸ɺԻ੠ͷ

εϖΫτϧΛ̐ଳҬʹ෼ׂ͢ΔΑ͏ͳಛ௃Λ͍࣋ͬͯΔɻ͜ͷ͜ͱ͸ɺઌʹड़΂ͨ̐ଳҬͷ νϟϯωϧϘίʔμԻ੠͕ߴ͍໌ྎ౓Λ΋ͭ͜ͱͱԿΒ͔ͷؔ࿈͕͋Δ͜ͱΛ͏͔͕ΘͤΔɻ Ellermeier et al. (2015)͸ɺUeda and Nakajima (2017)͕औΓग़ͨ̏ͭ͠ͳ͍̐ͭ͠ͷύϫʔ εϖΫτϧҼࢠʹΑͬͯ෼ׂ͞ΕΔ̐ଳҬʹैͬͯυΠπޠɾ೔ຊޠͷ̐ଳҬࡶԻۦಈԻ੠Λ

߹੒͠ɺௌऔ࣮ݧʹΑͬͯ͜ͷ̐ଳҬࡶԻۦಈԻ੠͕ߴ͍໌ྎ౓Λ΋ͭ͜ͱΛ͔֬Ί͍ͯΔɻ

͜ͷݚڀ͸ύϫʔεϖΫτϧҼࢠ͕΋ͭԻ੠஌֮ͷख͕͔Γʹ͍ͭͯߟ࡯͢ΔͨΊͷॏཁͳ ใࠂͰ͋Δɻ͔͠͠ͳ͕ΒɺΑΓ௚઀తͳߟ࡯Λߦ͏ͨΊʹ͸ɺύϫʔεϖΫτϧҼࢠ͕ද ݱ͠͏Δ৘ใ͚ͩΛ࣋ͭԻ੠Λ߹੒͠ɺͦΕΛ༻͍ͨௌऔ࣮ݧΛߦ͏͜ͱ͕ඞཁͰ͋Ζ͏ɻ Zahorian and Rothenberg (1981)ʹ͓͍ͯͦͷΑ͏ͳࢼΈ͕ͳ͞Ε͍ͯΔɻ൴Β͸ Plomp et

al. (1967)͕ߦͬͨओ੒෼෼ੳͱಉ༷ͷํ๏ͰऔΓग़͞Εͨओ੒෼͔ΒԻ੠Λ࠶߹੒͠ɺͦͷ

Ի੠ͷ໌ྎ౓Λଌఆ͍ͯ͠Δ͕ɺ൴Βͷݚڀ͸෼ੳʹ͓͚Δ࠷దͳ৚݅ͷ୳ࡧʹओ؟͕ஔ͔Ε

ͯ͋ΓɺԻ੠஌֮ʹ͓͍ͯओ੒෼͕΋ͭҙຯʹ͍ͭͯͷߟ࡯͸े෼ʹͳ͞Ε͍ͯͳ͔ͬͨɻ

1.1.4 ݴޠͷϦζϜͱ໐Իੑ

Ի੠஌֮ͷݚڀ͸୯ಠͰൃ࿩͞ΕͨԻૉɺԻઅɺ୯ޠͷ஌֮Λௐ΂ΔྖҬͱɺձ࿩࣌ͷΑ͏

ʹ࿈ଓతʹൃ࿩͞ΕͨԻ੠ͷ஌֮Λௐ΂ΔྖҬʹ෼͔ΕΔͱ࠷ॳʹड़΂ͨɻ೔ৗͷதͰൃ࿩͞

ΕΔԻ੠͸Իڹతʹ͸੾Ε໨ͷͳ͍࿈ଓମͰ͋Δɻ͜ͷ࿈ଓతͳԻΛௌ͍ͯݴޠͱͯ͠ਖ਼͘͠

ೝࣝ͢ΔͨΊʹ͸ɺ੾Ε໨ͷͳ͍Ի੠Λ͋Δ୯Ґʹ෼અ͢Δͱ͍͏ॲཧ͕ߦΘΕͳ͚Ε͹ͳ Βͳ͍ɻͰ͸ɺͲͷΑ͏ͳ୯Ґʹ෼અ͞ΕͯԻ੠͸஌֮͞Ε͍ͯΔͷͰ͋Ζ͏͔ɻ͜͜Ͱ͸ɺ ݴޠͷϦζϜʹয఺Λ౰ͯͯ͜ͷ໰୊ʹ͍ͭͯऔΓ্͛Δɻ

ݴޠʹ͸ϦζϜ͕͋Δɻ୯ޠɾจઅɾจʹΑͬͯϦζϜ͸֊૚తʹ࡞ΒΕɺϦζϜ͸࿩ऀͷ ײ৘Λ఻͑Δ໨త΍ɺಛఆͷޠΛڧௐ͢Δ໨తͰ༻͍ΒΕΔ͜ͱ΋͋Δ(Handel, 1989)ɻҟͳ ΔݴޠͷԻ੠Ͳ͏͠Λௌ͖ൺ΂ͯΈΕ͹ɺͦͷϦζϜ͕ҟͳΔ͜ͱʹؾͮͩ͘Ζ͏ɻ࣮ࡍʹݴ ޠ͸ͦͷϦζϜߏ଄ʹΑ͍͔ͬͯͭ͘ͷ୅දతͳάϧʔϓʹ෼ྨ͞ΕΔɻRamus, Nespor, and

(18)

Mehler (1999)͸༷ʑͳݴޠͷԻ੠Λ෼ੳ͠ɺ฼Իͷ۠ؒͷׂ߹ͱ̍จ಺ͷࢠԻͷ۠ؒͷׂ߹

ͷඪ४ภࠩͱͰද͞ΕΔฏ໘ʹ֤ݴޠΛ഑ஔ͢Δͱɺ̏ͭͷάϧʔϓʹ෼͔Εͯ഑ஔ͞ΕΔ

͜ͱΛࣔͨ͠ɻ͜ͷ̏ͭͷάϧʔϓ͸ݴޠͷϦζϜߏ଄ͷ୅දతͳάϧʔϓͱ͞ΕΔɺετ ϨελΠϛϯάݴޠ(stress-timed language)ɺԻઅλΠϛϯάݴޠ(syllable-timed language)ɺ ϞʔϥλΠϛϯάݴޠ(mora-timed language)ʹͦΕͧΕରԠ͍ͯ͠Δɻྫ͑͹ӳޠɾυΠπ ޠ͸ετϨελΠϛϯάݴޠɺϑϥϯεޠɾΠλϦΞޠ͸ԻઅλΠϛϯάݴޠ(Ladefoged &

Johnson, 2011)ɺͦͯ͠೔ຊޠɾλϛϧޠ͸ϞʔϥλΠϛϯάݴޠͰ͋Δ(Port, Dalby, & Oʟ Dell, 1987; Ramus et al., 1999)ɻݴޠʹ͓͚ΔϦζϜͷ໾ׂͷҰͭ͸஌֮ͷ୯ҐΛܗ੒͢Δ͜

ͱͰ͋ΔͱݴΘΕ͍ͯΔ(Cutler, 1994)ɻ஌࣮֮ݧʹΑͬͯɺӳޠԻ੠͕ετϨεͷ୯ҐͰɺϑ ϥϯεޠԻ੠͕Իઅͷ୯ҐͰ஌֮͞Ε͍ͯΔ͜ͱΛ(Cutler, Mehler, Norris, & Segui, 1986)ɺ

೔ຊޠԻ੠ʹ͓͍ͯ͸ϞʔϥΛ୯Ґʹͯ͠஌֮͞Ε͍ͯΔ͜ͱΛ(Otake, Hatano, Cutler, &

Mehler, 1993)ࣔ͢σʔλ͕ಘΒΕ͍ͯΔɻ

ݴޠͷϦζϜͱύϫʔεϖΫτϧҼࢠʹؔ࿈͕͋Δ͜ͱΛࣔ͢ݚڀ͕͋ΔɻYamashita et

al. (2013)͸ɺӳޠͱ೔ຊޠͷͦΕͧΕͷݴޠ؀ڥԼͰҭͯΒΕͨೕ༮ࣇ͕ࣗવʹൃ੠ͨ͠੠

Λܧଓతʹ࿥Ի͠ɺೕ༮ࣇͷ੒௕ͷաఔʹ͓͍ͯԻ੠ͷԻڹతಛ௃ʹͲͷΑ͏ͳมԽ͕ݟΒ ΕΔͷ͔Λ؍࡯ͨ͠ɻ൴ঁΒ͸݄ྸ15ɺ20ɺ24͔݄ͷ̏ͭͷ࣌ظͷԻ੠ʹରͯ͠ Ueda and

Nakajima (2017)ͱಉ͡ํ๏ͰҼࢠ෼ੳΛߦ͍ɺ݄ྸ͕ߴ͍ೕ༮ࣇͷԻ੠΄ͲɺύϫʔεϖΫ

τϧҼࢠͷύλʔϯ͕੒ਓͷ΋ͷʹ͍ۙ͜ͱΛݟ͚ͭͨɻ͞ΒʹɺҼࢠ෼ੳͰ̏ҼࢠΛऔΓग़

ͨ͠͏ͪͷҰͭͰ͋Δ1100 Hz෇ۙͷதଳҬʹେ͖͍ҼࢠෛՙྔΛ΋ͭҼࢠͷҼࢠಘ఺ʹͭ

͍ͯɺͦͷࣗݾ૬ؔؔ਺ΛٻΊΔ͜ͱͰͦͷҼࢠಘ఺ͷ࣌ؒมಈʹ͓͍ͯϦζϜύλʔϯͷ Α͏ͳ΋ͷ͕ݟΒΕΔͷ͔Λௐ΂ͨɻࣗݾ૬ؔؔ਺ͷ஋ʹ͸͖ͬΓͱͨ͠ϐʔΫ͕ݱΕͨͱ͖

ʹɺͦͷϐʔΫ͕Ͱ͖Δ࣌ؒΛִ࣌ؒؒͱ͢ΔϦζϜ͕ܗ੒͞Ε͍ͯΔͱߟ͑Δ͜ͱ͕Ͱ͖

Δɻ͜ͷ෼ੳʹΑͬͯɺ݄ྸͷߴ͍ೕ༮ࣇͷԻ੠ͷϦζϜ͸੒ਓͷԻ੠ͷϦζϜʹ͍ۙ͜ͱ͕

෼͔ͬͨɻ

1100 Hz෇ۙͷଳҬʹେ͖͍ҼࢠෛՙྔΛ΋ͭύϫʔεϖΫτϧҼࢠ͸ݴޠͷϦζϜΛௐ

΂Δ͜ͱʹར༻Ͱ͖Δ͜ͱΛ͕ࣔͨ͠ɺ͜ͷҼࢠʹ͍ͭͯ͞Βʹৄ͘͠෼ੳͨ͠ Nakajima, Ueda, Fujimaru, Motomura, and Ohsaka (2017)ͷݚڀʹ͍ͭͯ৮ΕΔɻNakajima et al. (2017)

͸ΠΪϦεӳޠԻ੠ʹରͯ͠ Ueda and Nakajima (2017)Ͱ༻͍ΒΕͨԻ੠ͷεϖΫτϧߏ଄

ʹର͢ΔҼࢠ෼ੳΛ༻͍ͯ̏ͭͷύϫʔεϖΫτϧҼࢠΛऔΓग़͠ɺ֤ԻૉͷҼࢠಘ఺Λ؍

࡯ͨ͠ɻ൴Β͸֤ԻૉΛҼࢠಘ఺ʹ͕ͨͬͯ̏͠Ҽࢠͷ্ۭؒʹ഑ஔͤ͞Δͱɺ֤Իૉ͕͋

(19)

Δۂઢ্Λɺ໐Իੑ(sonority) ͷई౓ͷॱʹै͏Α͏ʹ෼෍͢Δ܏޲͕͋Δ͜ͱΛݟ͚ͭͨɻ

͞Βʹ͔ͦ͜Βɺ1100 HzपลͷதଳҬʹ͓͍ͯҼࢠෛՙྔ͕େ͖͍ύϫʔεϖΫτϧҼࢠʹ ໐Իੑͷई౓ͱਖ਼ͷ૬͕ؔ͋Δ͜ͱ͕ɺͦͯ͠3300 HzҎ্ͷߴଳҬʹ͓͍ͯҼࢠෛՙྔ͕

େ͖͍ύϫʔεϖΫτϧҼࢠʹ໐Իੑͷई౓ͱෛͷ૬͕ؔ͋Δ͜ͱΛݟ͚ͭͨɻ໐Իੑͱ͸ɺ

ͦΕͧΕͷԻૉʹ͍ͭͯɺͦΕΒΛͲΕ͚ͩେ͖͘ڹ͔ͤͯൃ࿩Ͱ͖Δ͔Λࣔ͢ॱংई౓Ͱ

͋Γɺݴޠֶ΍Ի੠ֶͷݚڀऀΒʹΑͬͯఏএ͞Εͨ΋ͷͰ͋Δ(Selkirk, 1984; Harris, 1994;

Spencer, 1996)ɻde Saussure (1959)͸ɺൃ࿩ͷࡍʹͲΕ͘Β͍ௐԻث׭͕։͍͍ͯΔ͔ɺͦ

ΕʹΑͬͯͲΕ͚ͩԻ͕ڹ͔͘ͱ͍͏؍఺ͰԻૉΛ։ޱ౓ (aperture)ͱ͍͏ॱংई౓Ͱ෼ྨ

͍ͯ͠Δɻ։ޱ౓ͱݺশ͞Ε͍ͯΔ͕ɺௐԻͱௌ֮ͱ͕੾Γ཭ͤͳ͍΋ͷͱ͍͏ߟ͑ʹج͍ͮ

͓ͯΓɺԻͷௌ͑͜ͱͷؔ܎ʹॏ͖Λஔ͍ͯߟ࡯͕ਐΊΒΕ͍ͯΔɻ։ޱ౓΋໐Իੑͱಉ༷ͷ

΋ͷͰ͋Δͱߟ͑Δ͜ͱ͕Ͱ͖ΔɻSpencer (1996)ʹΑΔ໐Իੑͷई౓Ͱ͸ɺ฼Իɺ౉ΓԻɺ

ྲྀԻɺඓԻɺຎࡲԻɾഁࡲԻɺഁ྾Իͷॱʹ໐Իੑ͕௿͘ͳΔͱ͍ͯ͠ΔɻԻઅ͸Իૉ͕࿈݁

͞ΕΔ͜ͱʹΑͬͯߏ੒͞ΕΔ͕ɺجຊతʹ໐Իੑ͕௿͍Իૉ͔Βߴ͍Իૉ΁ͱͭͳ͕Γɺͦ

ͯ͠·ͨ௿͍Իૉʹͭͳ͕ΔΑ͏ʹͳ͍ͬͯΔɻ͜Ε͸໐Իੑ࿈ଓݪཧ(sonority sequencing

principle; Rahilly, 2016)ͱݺ͹ΕΔ΋ͷͰɺ͜ͷنଇʹैͬͯԻૉ͕࿈ͳΔͱɺ໐Իੑͷࢁ͕

Ͱ͖Δ৔ॴʹԻઅͷ͕֩ܗ੒͞ΕΔɻNakajima et al. (2017)ͷݚڀͷಛච͢΂͖఺͸ɺ໐Ի

ੑͷਫ਼ਆ෺ཧֶత࣮ମΛఏҊͨ͜͠ͱͰ͋Γɺ͜ͷํ๏Ͱ໐ԻੑΛఆٛ͢Ε͹ɺ୯ޠ stopͷ Α͏ͳӳޠʹ͓͍ͯසग़͢ΔɺຎࡲԻ/s/ͱഁ྾Ի/t/ͷ಄ࢠԻ࿈݁ʹ͓͍ͯɺ/s/͕Իઅͷ֩

ͱ͸ͳΒͳ͍͜ͱΛ໐Իੑ࿈ଓݪཧʹໃ६ͤͣʹઆ໌͢Δ͜ͱ͕Ͱ͖Δɻ

Ի੠Λௌ͍ͨͱ͖ʹײͥΒΕΔϦζϜ͸ڧऑͷཁૉ͕࣌ؒతʹنଇੑΛ΋ͬͯฒΜͰ͍ΔͷΛ

஌֮͢Δ͜ͱͰܗ੒͞ΕΔɻ͜ͷԻ੠ͷڧऑͷཁૉ͕໐ԻੑͰ͋Δͱߟ͑ΒΕ͍ͯΔ(Handel, 1989)ɻGalves, Garcia, Duarte, and Galves (2002)΋໐Իੑ͕ݴޠͷϦζϜͱؔ࿈͕͋Δ͜ͱ ΛҟͳΔϦζϜߏ଄Λ΋ͭݴޠͷԻ੠ΛԻڹతʹ෼ੳ͢Δ͜ͱͰ͔֬Ί͍ͯΔɻ ݴޠͷϦζϜ ΛͱΒ͑Δ͜ͱ͕Ի੠ͷ஌֮ʹॏཁͰ͋Δ͜ͱ͔Βɺ໐Իੑ͕Ի੠ͷ஌֮ʹͲͷΑ͏ʹӨڹΛ ༩͑Δͷ͔ʹ͍ͭͯௐ΂Δ͜ͱ΋·ͨॏཁͰ͋Ζ͏ɻUeda and Nakajima (2017)ɺNakajima

et al. (2017)ͷݚڀʹΑͬͯ໐Իੑ͕ύϫʔεϖΫτϧҼࢠͱ͍͏ଌఆͰ͖Δ΋ͷͰͱΒ͑Δ

͜ͱ͕Ͱ͖ΔΑ͏ʹͳͬͨɻΑͬͯύϫʔεϖΫτϧҼࢠ͔Β௚઀Ի੠Λ࠶߹੒͢Ε͹ɺԻ੠

஌֮ͱ໐Իੑͷؔ܎Λௐ΂Δௌऔ࣮ݧΛߦ͏͜ͱ͕Ͱ͖Δɻ

(20)

1.2 ຊ࿦จͷ໨త

ຊ࿦จͰ͸ɺԻ੠ͷྟքଳҬ͝ͱͷύϫʔมಈΛߏ੒͢ΔύϫʔεϖΫτϧҼࢠʹ͍ͭͯɺ Ի੠஌֮ʹ͓͚Δͦͷ໾ׂΛௌऔ࣮ݧʹΑͬͯௐ΂Δ͜ͱΛ໨తͱ͢Δɻ͜ΕΛ࣮ݱ͢ΔͨΊ ʹɺύϫʔεϖΫτϧҼࢠ͔ΒԻ੠Λ࠶߹੒͢Δख๏Λཱ֬͢Δɻ߹੒Ի੠Λ༻͍ͨԻ੠ͷ໌

ྎ౓Λଌఆ͢Δௌऔ࣮ݧͱɺ౷ܭతख๏ʹΑΔԻ੠ͷεϖΫτϧͷߏ଄෼ੳͱΛ݁ͼ͚ͭΔݚ ڀʹҐஔ͚ͮΒΕΔɻ

1.3 ຊ࿦จͷߏ੒

ୈ̍ষͰ͸ɺຊݚڀͷഎܠͱͯ͠ɺௌ֮ܥ຤ধʹΑͬͯಘΒΕΔԻ੠ͷεϖΫτϧදݱ͕Ի

੠ͷ஌֮ͷࡍʹͲͷΑ͏ʹར༻͞Ε͍ͯΔͷ͔ʹ͍ͭͯɺௌऔ࣮ݧΛ௨ͯ͠ௐ΂ͨઌߦݚڀ͓

ΑͼԻ੠ͷ౷ܭత෼ੳΛ௨ͯ͠ௐ΂ͨઌߦݚڀΛ঺հͨ͠ɻͦͷதͰ໰୊఺Λ੔ཧ͠ɺຊݚڀ ͷ໨తΛࣔͨ͠ɻ

Ҏ߱ɺୈ̎ষ͔Βୈ̐ষʹ͓͍ͯɺຊݚڀͰߦͬͨ2ͭͷ෼ੳ͓Αͼ4ͭͷ࣮ݧʹ͍ͭͯใ ࠂ͢Δɻୈ̎ষͷ෼ੳ̍Ͱ͸ Ueda and Nakajima (2017)ʹΑΔԻ੠ͷྟքଳҬ͝ͱͷύϫʔ มಈʹର͢ΔҼࢠ෼ੳʹΑͬͯಘΒΕΔύϫʔεϖΫτϧҼࢠ͔ΒɺྟքଳҬ͝ͱͷύϫʔม ಈΛ࠶ߏ੒͠ɺௌऔ࣮ݧʹ༻͍ΔͨΊͷܹࢗԻΛ࡞੒͢Δͷʹదͨ͠Ҽࢠ෼ੳ๏ΛఏҊ͢Δɻ

͜ͷҼࢠ෼ੳ๏ʹΑͬͯ೔ຊޠɾΠΪϦεӳޠɾதࠃޠ(ී௨࿩)ͷԻ੠Λ෼ੳ͠ɺಘΒΕͨ

ύϫʔεϖΫτϧҼࢠ͕ Ueda and Nakajima (2017)ͷ෼ੳͷ΋ͷͱಉ౳ͷҼࢠͰ͋Δͷ͔Λ

֬ೝ͢Δɻଓ࣮͘ݧ̍Ͱ͸ɺύϫʔεϖΫτϧҼࢠʹΑͬͯදݱͰ͖ΔԻ੠ͷύϫʔεϖΫτ ϧͷ࣌ؒมԽͷ৘ใʹΑͬͯɺ೔ຊޠԻ੠͕Ͳͷఔ౓ਖ਼֬ʹ఻͑ΒΕΔͷ͔Λௌऔ࣮ݧͰௐ΂

Δɻ͜ͷ࣮ݧʹΑͬͯύϫʔεϖΫτϧҼࢠΛ͍ͭ͘·Ͱ༻͍Ε͹ɺे෼ʹ໌ྎͳԻ੠Λ߹੒

͢Δ͜ͱ͕Ͱ͖Δͷ͔Λ͔֬ΊΔɻୈ̏ষͰ͸ୈ̎ষͷύϫʔεϖΫτϧҼࢠΛ༻͍ͨԻ੠

ͷ࠶߹੒ͷࡍʹى͖Δ໰୊఺ʹ஫໨͠ɺ͜ͷ໰୊Λճආ͢Δํ๏ͱͯ͠ύϫʔεϖΫτϧҼ ࢠͷ௚ަੑΛҡ࣋ͨ͠··ඇෛ஋Խͨ͠΋ͷʹमਖ਼͢Δํ๏ΛఏҊ͢Δɻ෼ੳ̎ͱͯ͠ɺύ ϫʔεϖΫτϧҼࢠʹΑΔɺԻ੠ͷύϫʔεϖΫτϧมԽͷઆ໌཰͕͜ͷमਖ਼ʹΑͬͯͲΕͩ

͚ӨڹΛड͚Δͷ͔Λௐ΂Δɻ࣮ݧ̎Ͱ͸࣮ݧ̍ͱಉ༷ͷํ๏Λ༻͍ͯɺ͜ͷඇෛ஋ԽΛߦͬ

ͨύϫʔεϖΫτϧҼࢠΛ༻͍ͯ߹੒͞ΕͨԻ੠ͷ໌ྎ౓Λଌఆ͢Δɻ࣮ݧ̍ͱ࣮ݧ̎ͷ݁

ՌΛൺֱ͢Δ͜ͱʹΑͬͯɺୈ̎ষʹ͓͚ΔԻ੠ͷ࠶߹੒ͷࡍʹੜ͍ͯͨ͡໰୊͕ຊݚڀͷ໨

తΛୡ੒͢Δ্Ͱॏཁͳ໰୊Ͱ͋Δͷ͔Λݕ౼͢Δɻୈ̐ষͰ͸ୈ̎ষ͓Αͼୈ̏ষͰ෼͔ͬ

(21)

ͨԻ੠ͷ໌ྎͳ஌֮ʹ͓͍ͯॏཁͱͳΔύϫʔεϖΫτϧҼࢠʹ͍ͭͯɺݸʑͷҼࢠͷ໾ׂʹ

஫໨͢ΔɻԻ੠ΛύϫʔεϖΫτϧҼࢠ͔Β࠶߹੒͢Δࡍʹɺ͍͔ͭ͘ͷύϫʔεϖΫτϧҼ ࢠʹΑͬͯ༩͑ΒΕΔύϫʔεϖΫτϧͷ࣌ؒมԽͷ৘ใΛऔΓআ͖ɺ࠶߹੒͞ΕͨԻ੠ͷ໌

ྎ౓͕ͲΕ͚ͩ௿Լ͢Δͷ͔Λௐ΂ΔɻऔΓআ͘Ҽࢠ͕ҟͳΔͱɺ໌ྎ౓͕ͲΕ͚ͩҟͳΔͷ

͔Λൺֱ͢Δ͜ͱͰɺҼࢠͷ΋ͭݸʑͷ໾ׂʹ͍ͭͯߟ࡯͢Δɻ

ୈ̑ষͰ͸૯߹ߟ࡯Λߦ͏ɻ·ͣɺୈ̎ষ͔Β̐ষʹ͓͍ͯߦͬͨ෼ੳ͓Αͼ࣮ݧͷ݁ՌΛ

·ͱΊɺύϫʔεϖΫτϧҼࢠ͕Ի੠ͷ஌֮ʹͲͷΑ͏ͳ໾ׂΛ΋ͭͷ͔ʹ͍ͭͯ݁࿦Λड़

΂Δɻ࣍ʹಋ͔Εͨ݁࿦͕ݚڀ࢙ͷதͰͲͷΑ͏ʹҐஔ͚ͮΒΕΔͷ͔ɺ·ͨ͸ઌߦݚڀʹର

ͯ͠ͲͷΑ͏ͳ৽͍͠ղऍΛ༩͑Δͷ͔ʹ͍ͭͯߟ࡯͢ΔɻͦͷதͰɺΘ͔ͣͳଳҬ਺ͷνϟ ϯωϧϘίʔμԻ੠Ͱͳͥ໌ྎʹ಺༰Λ஌֮͢Δ͜ͱ͕Ͱ͖Δ͔ʹ͍ͭͯɺຊݚڀͷ݁࿦͔Β આ໌ΛࢼΈΔɻ࠷ޙʹݚڀͷ໰୊఺ʹ͍ͭͯ৮Εɺࠓޙͷల๬Λड़΂Δɻ

(22)

ୈ 2 ^ষ จԻ੠ͷ໌ྎͳ஌֮ʹཁ͢Δύϫʔ εϖΫτϧҼࢠͷݸ਺

2.1 ୈ̎ষͷ໨త

զʑ͕Ի੠Λௌऔ͢Δͱɺௌ֮ͷप೾਺෼ੳػೳʹΑͬͯௌ֮ܥ຤ধʹ͓͍ͯԻ੠ͷεϖΫ τϧදݱ͕ಘΒΕ͍ͯΔɻྟքଳҬϑΟϧλΛ༻͍Δ͜ͱͰɺௌ֮ܥ຤ধʹ͓͚ΔԻ੠ͷεϖ ΫτϧදݱΛ໛ٖ͢Δ͜ͱ͕Ͱ͖ΔɻUeda and Nakajima (2017)͕ಋग़ͨ͠ύϫʔεϖΫτ ϧҼࢠ͸20ݸͷྟքଳҬϑΟϧλʹΑͬͯಘΒΕͨԻ੠ͷύϫʔεϖΫτϧΛΑΓগͳ͍ݸ

਺ͷҼࢠͷઢܗ݁߹ʹΑͬͯۙࣅ͢Δ΋ͷͰ͋Δɻ͢ͳΘͪɺऔΓग़͢Ҽࢠ਺͕20ݸʹ͍ۙ

΄ͲྟքଳҬϑΟϧλͷग़ྗΛ஧࣮ʹ࠶ݱͰ͖ΔɻUeda and Nakajima (2017)͕෼ੳର৅ͱ

ͨ̔͠ݴޠؒͰڞ௨ͨ͠Ҽࢠߏ଄͕ಘΒΕͨͷ͸̐Ҽࢠ·ͰͰ͋ͬͨɻ൴ΒͷҼࢠ෼ੳ͸ओ੒

෼෼ੳΛجૅͱ͍ͯ͠Δ͜ͱ͔Βɺ͜ͷ෼ੳ݁Ռ͸ɺԻ੠ͷύϫʔεϖΫτϧߏ଄ͷ͏ͪͷ ओཁͳಛ௃Λߏ੒͢ΔͨΊͷ4Ҽࢠ͕֤ݴޠʹ͓͍ͯڞ௨͍ͯ͠Δͱ͍͏͜ͱΛ͍ࣔͯ͠Δɻ

ͦΕͰ͸ɺݴޠΛ௒͑ͯڞ௨͢Δಛ௃Λ࣋ͬͨԻ੠ͷύϫʔεϖΫτϧҼࢠ͸Ի੠Λ໌ྎʹ஌

֮͢ΔͨΊʹͲΕ͚ͩͷ৘ใΛ༩͑͏ΔͷͩΖ͏͔ɻ

ຊষͰ͸ɺԻ੠Λ࠶߹੒͢Δͷʹదͨ͠ύϫʔεϖΫτϧҼࢠΛಘΔ͜ͱ͕Ͱ͖Δ৽͍͠ओ

੒෼෼ੳɺʮى఺Ҡಈओ੒෼෼ੳ (origin-shifted principal component analysis)ʯΛఏҊ͢Δɻ

෼ੳ̍ͱͯ͠ɺఏҊख๏ʹΑͬͯԻ੠Λ෼ੳͨ͠৔߹ɺઌߦݚڀͰಘΒΕͨҼࢠͱಉ౳ͷҼࢠ

͕ಘΒΕΔ͔Λ֬ೝ͢Δɻ࣍ʹ࣮ݧ̍ͱͯ͠ɺύϫʔεϖΫτϧҼࢠ͔Β೔ຊޠԻ੠ΛࡶԻۦ ಈԻ੠ͱͯ͠࠶߹੒ͨ࣌͠ɺҼࢠΛ͍ͭ͘·Ͱ༻͍Ε͹࠶߹੒͞ΕͨࡶԻۦಈԻ੠Λे෼໌ྎ

ʹௌ͖औΔ͜ͱ͕Ͱ͖Δͷ͔Λௐ΂Δɻ

(23)

2.2 ෼ੳ̍ɿ ى఺Ҡಈओ੒෼෼ੳʹΑΔύϫʔεϖΫτϧҼࢠ ͷநग़

ैདྷͷओ੒෼෼ੳʹΑͬͯಘΒΕΔύϫʔεϖΫτϧҼࢠ͔ΒԻ੠Λ࠶߹੒͢Δ৔߹ɺޙड़

͢ΔఆৗࡶԻ੒෼͕ൃੜ͢Δͱ͍͏໰୊͕͋Γɺௌऔ࣮ݧʹ༻͍Δͷ͕ෆద੾Ͱ͋Δͱߟ͑Β ΕΔɻͦ͜Ͱओ੒෼෼ੳͷม๏ΛఏҊ͠ɺ͜ͷ໰୊Λճආ͢Δ͜ͱͱͨ͠ɻ͜ͷઅͰ͸ɺ৽͠

͘ఏҊ͢Δʮى఺Ҡಈओ੒෼෼ੳʯΛ௨ͯ͠ಘΒΕΔύϫʔεϖΫτϧҼࢠͱैདྷͷओ੒෼෼

ੳΛ௨ͯ͠ಘΒΕΔύϫʔεϖΫτϧҼࢠͱΛൺֱ͠ɺ෼ੳ๏ͷมߋ͕ɺ݁Ռʹରͯ͠ຊ࣭త ͳӨڹΛ༩͍͑ͯͳ͍͔Λ͔֬ΊΔɻ

2.2.1 ෼ੳࢼྉ

NTT-ATࣾͷʮଟݴޠԻ੠σʔλϕʔε2002 (NTT-AT, 2002)ʯʹσΟδλϧऩ࿥(16-bit

ྔࢠԽɺ16000 Hz αϯϓϦϯά)͞Εͨɺ೔ຊޠɺΠΪϦεӳޠɺதࠃޠ(ී௨࿩)Ի੠Λ෼

ੳࢼྉͱͯ͠༻͍ͨɻ೔ຊޠԻ੠ɺΠΪϦεӳޠԻ੠͸ͦΕͧΕ200จ͔ΒͳΓɺ֤ݴޠͷ฼

ޠ࿩ऀͰ͋Δஉੑ໊͕̑ͦΕͧΕͷจΛൃ࿩ͨ͠ɻதࠃޠ(ී௨࿩)ʹ͍ͭͯ͸ɺ78จ͔Βͳ Γɺ฼ޠ࿩ऀͰ͋Δஉੑ໊͕̑ͦΕͧΕͷจΛൃ࿩ͨ͠ɻ֤จ͸ฏۉ̎sఔ౓ͷ௕͞Ͱൃ࿩͞

Ε͍ͯΔɻ͢΂ͯͷ෼ੳର৅ͷԻ੠ͷ૯ൃ࿩࣌ؒ͸ɺ೔ຊޠɺΠΪϦεӳޠɺதࠃޠ(ී௨࿩) ͷॱʹɺ2484 sɺ1979 sɺ870 sͰ͋ͬͨɻ͜Ε͸ɺ҆ఆͨ͠෼ੳ݁Ռ͕ಘΒΕΔͷʹඞཁͳ

௕͞Ͱ͋Δͱ͞ΕΔ30 s (Li, Hughes, & House, 1969; Zahorian & Rothenberg, 1981)Λे෼

ʹ௒͍͑ͯΔɻԻ੠ͷฏۉجຊप೾਺͸೔ຊޠɺΠΪϦεӳޠɺதࠃޠ(ී௨࿩)ͰͦΕͧΕ 136 HzʢSD = 31 Hzʣɺ126 HzʢSD = 30 Hzʣɺ164 HzʢSD = 38 HzʣͰ͋ͬͨɻ͜ΕΒͷ

̏ͭͷݴޠ͸ɺҟͳΔݴޠάϧʔϓͷ୅දͱͯ͠ɺUeda and Nakajima (2017)Ͱ෼ੳ͞Εͨ

̔ݴޠͷத͔Βબग़ͨ͠ɻ೔ຊޠɾΠΪϦεӳޠɾதࠃޠ(ී௨࿩)͸ޓ͍ʹҟͳΔݴޠϦζ ϜΛ͓࣋ͬͯΓɺΠΪϦεӳޠ͸ετϨελΠϛϯάݴޠɺ೔ຊޠ͸ϞʔϥλΠϛϯάݴޠɺ தࠃޠ͸ԻઅλΠϛϯάݴޠͰ͋Δ(Cutler, 1994; Ramus et al., 1999)ɻUeda and Nakajima

(2017)ͷ෼ੳͰ͸ɺஉੑ࿩ऀ͚ͩͰͳ͘ঁੑ࿩ऀͷԻ੠΋෼ੳʹ༻͍ΒΕ͍ͯͨɻຊ࿦จͰ

͸ɺௌऔ࣮ݧͰஉੑ࿩ऀͷԻ੠ΛݪԻ੠ʹ༻͍ͨɻஉੑ࿩ऀͷԻ੠͸ঁੑ࿩ऀͷԻ੠ΑΓ΋ج ຊप೾਺͕௿͍ɻجຊप೾਺͕௿͍Ի੠͸ύϫʔεϖΫτϧͷแབྷܗঢ়ͷ৘ใ͕औΓग़͠΍͢

͍ɻͦͷͨΊɺஉੑ࿩ऀͷԻ੠Λௌऔ࣮ݧʹ༻͍ΔݪԻ੠ͱͨ͠ɻௌऔ࣮ݧͰ༻͍Δύϫʔε ϖΫτϧҼࢠΛಘΔͨΊʹɺஉੑ࿩ऀͷԻ੠ͷΈΛ෼ੳ͢Δ͜ͱʹͨ͠ɻ

(24)

2.2.2 खଓ͖

෼ੳखଓ͖Λਤ2.1ʹࣔ͢ɻ෼ੳ͸֤Ի੠৴߸Λॲཧͯ͠ྟքଳҬ͝ͱͷύϫʔมಈΛಘΔ ෦෼ͱɺྟքଳҬ͝ͱͷύϫʔมಈΛى఺Ҡಈओ੒෼෼ੳʹ͔͚ͯύϫʔεϖΫτϧҼࢠΛಘ Δ෦෼ͱʹ෼͚ΒΕΔɻ

Signal Processing

…

File 1 File 2 File 200

Origin-shifted principal component analysis

&

Varimax rotation 1

2 20

Power fluctuations in each critical band

Spectral change factors Speech signals 0.5 s

0.5 s

75 150 250 350 450 570 700 840 1000 1170 1370 1600 1850 2150 2500 2900 3400 4000 4800 5800 Center frequency (Hz)

ਤ 2.1෼ੳखଓ͖ͷྲྀΕਤɻσΟδλϧ࿥Ի͞ΕͨݪԻ੠͸৴߸ॲཧΛܦͯ20ͷྟքଳҬ͝

ͱͷύϫʔมಈͱͳΔɻ20ͷྟքଳҬ͝ͱͷύϫʔมಈΛى఺Ҡಈओ੒෼෼ੳͱόϦϚοΫ εճసʹ͔͚ͯɺύϫʔεϖΫτϧҼࢠ͕ಘΒΕΔɻ

(25)

·ͣ͸ɺҼࢠ෼ੳʹ͔͚ΔσʔλͱͳΔྟքଳҬ͝ͱͷύϫʔมಈΛಘΔͨΊͷ෼ੳखଓ͖

ʹ͍ͭͯઆ໌͢Δɻ͜͜Ͱߦͬͨ৴߸ॲཧͷྲྀΕ͸ਤ2.2ͷΑ͏ʹ·ͱΊΒΕΔɻ

෼ੳର৅ͷԻ੠Λ20ͷྟքଳҬʹ෼ׂ͠ɺ֤ଳҬͷύϫʔมಈΛ̍ msִؒͰಘΔͨΊͷ ॲཧΛߦͬͨɻ෼ੳର৅Ի੠ͷ͋Δ࣌఺ͰͷύϫʔεϖΫτϧΛಘΔͨΊʹɺͦͷ࣌఺Λத৺

ͱ͢Δ30 msͷ۠ؒͷ࣌ؒ೾ܗΛ૭ؔ਺Ͱ෦෼తʹ੾Γऔͬͨɻ૭ؔ਺ʹ͸ϋϛϯά૭Λ༻͍

ͨɻ࣍ʹߴ଎ϑʔϦΤม׵Λߦ͍ɺϋϛϯά૭Ͱ੾ΓऔΒΕͨ30 ms෼ͷ୹࣌ؒ৴߸͔Βৼ ෯εϖΫτϧΛಘͨɻ͞Βʹͦͷৼ෯εϖΫτϧΛ̎৐ͯ͠ύϫʔεϖΫτϧΛࢉग़ͨ͠ɻ

͜͜·ͰͷॲཧͰಘΒΕͨύϫʔεϖΫτϧ͸ɺप೾਺্࣠Ͱύϫʔ͕ඍࡉʹมಈ͍ͯ͠

Δɻ͜ͷύϫʔεϖΫτϧͷඍࡉͳ੒෼͸ओʹ੠ଳৼಈʹىҼ͠ɺԻ੠ͷجຊप೾਺ʹؔ࿈͢

Δ੒෼Ͱ͋ΔɻҰํύϫʔεϖΫτϧͷแབྷͷಛ௃͸ൃ੠࣌ͷ੠ಓܗঢ়ʹΑܾͬͯ·ΔɻԻ੠

ͷύϫʔεϖΫτϧ͸੠ଳৼಈ(Իݯ)͕࡞Δप೾਺ಛੑͱ੠ಓܗঢ়(ϑΟϧλ)ʹΑܾͬͯ·

Δڞ໐ͷप೾਺ಛੑͱͷੵͰϞσϧԽ͞ΕΔɻ͜ͷߟ͑ํ͸ԻݯϑΟϧλཧ࿦(Fant, 1960)ͱ ݺ͹Ε͍ͯΔɻ੠ಓܗঢ়ʹىҼ͢ΔύϫʔεϖΫτϧแབྷͷಛ௃Λ෼ੳ͢Δ͜ͱ͕͜͜Ͱͷ໨

తͰ͋ΔɻUeda and Nakajima (2017)ͷ෼ੳ݁ՌͰ͸ɺ̐ҼࢠΛऔΓग़͢෼ੳʹ͓͍ͯɺத

৺प೾਺ͷ௿͍ྟքଳҬʹ͓͍ͯҼࢠෛՙྔ͕ྟքଳҬҰͭඈ͹͠Ͱେ͖ͳ஋ͱͳΔΑ͏ͳ Ҽࢠ͕औΓग़͞Ε͍ͯΔɻύϫʔεϖΫτϧͷඍࡉߏ଄ʹݟΒΕΔҰͭͻͱͭͷࢁͷִؒ͸੠

ଳৼಈͷपظɺ͢ͳΘͪجຊप೾਺ͱ͓͓ΉͶҰக͢Δɻجຊप೾਺͕ྟքଳҬ෯Λ௒͍͑ͯ

Δ৔߹͸ɺ͋ΔྟքଳҬͰύϫʔεϖΫτϧͷඍࡉߏ଄ͷࢁ͕Ͱ͖͍ͯΔͱ͖ʹͦͷྡͷྟք ଳҬͰ͸୩͕Ͱ͖ΔΑ͏ͳ͜ͱ͕͋Δɻ͜Ε͕ɺҼࢠෛՙྔ͕ྟքଳҬҰͭඈ͹͠Ͱେ͖ͳ஋

ͱͳͬͨݪҼͰ͋Ζ͏ɻͦ͜ͰຊݚڀͰ͸ύϫʔεϖΫτϧͷแབྷܗঢ়Λਪఆ͠ɺ͜ΕΛ෼ੳ

͢Δ͜ͱͱͨ͠ɻ ύϫʔεϖΫτϧแབྷΛਪఆ͢ΔͷʹԻڹ޻ֶͷ෼໺ͰΑ͘༻͍ΒΕ͍ͯ

ΔɺέϓετϥϜ෼ੳ(e.g., Rabiner & Schafer, 1978)Λ࠾༻ͨ͠ɻύϫʔεϖΫτϧͷର਺

ΛऔΓɺͦΕΛϑʔϦΤม׵͢Δ͜ͱͰಘΒΕΔέϓετϥϜʹ͸ͦͷߴ࣍ͷ੒෼(ߴέϑϨ ϯγ੒෼)ʹԻݯͷप೾਺ಛੑ͕ɺ௿࣍ͷ੒෼(௿έϑϨϯγ੒෼)ʹ੠ಓͷप೾਺ಛੑ͕ݱΕ

͍ͯΔ͜ͱ͕ԻݯϑΟϧλཧ࿦Λجʹࣔ͞Ε͍ͯΔɻ͋Δ஋Ҏ্ͷߴέϑϨϯγ੒෼Λέϓε τϥϜ͔ΒऔΓআ͍ͨ͏͑ͰɺύϫʔεϖΫτϧʹ໭͢͜ͱͰύϫʔεϖΫτϧͷৼ෯แབྷΛ ਪఆ͢Δ͜ͱ͕Ͱ͖Δɻ

(26)

4 6 8 10 12 14 16 18 20 22 24

2000 4000 6000

1000

0 3000 5000

8 10 12 14 16 18 20 22

2000 4000 6000

1000

0 3000 5000

30 ms

Fast fourier transform

Short-pass liftering

(5 ms)

Averaging in each critical band

Hamming window (30 ms)

Power (dB)Power (dB)Power (dB)

Frequency (Hz)

Center frequency (Hz)

75 150 250 350 450 570 700 840 1000 1170 1370 1600 1850 2150 2500 2900 3400 4000 4800 5800

8 10 12 14 16 18 20 22

2000 4000 6000

1000

0 3000 5000

Input: speech signal

Short segment

Power spectrum

Smoothed power spectrum

Critical band filter output

Output: power fluctuations

ਤ 2.2 Ի੠৴߸͔Β20ͷྟքଳҬ͝ͱͷύϫʔมಈΛಘΔͨΊͷ৴߸ॲཧखॱɻ

(27)

ͦ͜ͰຊݚڀͰ͸ɺέϑϨϯγ্࣠Ͱ̑ msҎ্ͷߴέϑϨϯγ੒෼Λআڈ(ϦϑλϦϯά)

ͨ͠ɻ͜ΕʹΑͬͯɺඍࡉͳߏ଄͕ฏ׈Խ͞ΕͨύϫʔεϖΫτϧΛಘͨɻ ϦϑλϦϯάʹ Αͬͯฏ׈Խ͞ΕͨύϫʔεϖΫτϧΛ20ͷྟքଳҬʹप೾਺෼ղ͠ɺ֤ྟքଳҬͰଳҬ಺ͷ ฏۉͷύϫʔ஋ΛٻΊͨɻҎ্ͷૢ࡞ͰԻ੠৴߸ͷ͋Δ࣌఺ʹ͓͚Δ20ͷྟքଳҬͷύϫʔ

஋͕ಘΒΕΔɻ͜ͷૢ࡞Λϋϛϯά૭ͷҐஔΛ̍ msͣͭͣΒͯ͠(ϑϨʔϜपظ̍msͰ)ߦ

͏͜ͱͰɺ20ͷྟքଳҬͷύϫʔมಈΛ̍ msִؒͰಘͨɻ20ͷྟքଳҬͷத৺प೾਺͓Α ͼःஅप೾਺͸ɺUeda and Nakajima (2017); Nakajima et al. (2017)ͱಉ͡৚݅Ͱߦ͏ͨΊ ʹɺ Zwicker and Terhardt (1980)Λࢀߟͱͨ͠ɻͨͩ͠ɺ50 HzҎԼͷଳҬ͸Ի੠ͱ͸ؔ࿈

͕খ͘͞ɺσʔλϕʔεऩ࿥ͷࡍʹআڈ͞Ε͍ͯΔͨΊɺୈ1൪໨ͷྟքଳҬ͕0–100 Hzͱ ͳ͍ͬͯΔͱ͜ΖΛɺ 50–100 Hz ʹมߋͨ͠ɻΑͬͯ50–6400 Hzͷൣғͷप೾਺ʹ20ͷྟ

քଳҬ͕഑ஔ͞Εͨ (ද2.1)ɻ෼ੳର৅ͷݪԻ੠͕΋ͱ΋ͱ8000 Hz·Ͱऩ࿥͞Ε͍ͯͨͷʹ ରͯ͠ɺ6400 Hzͷप೾਺੒෼·Ͱ͔͠෼ੳ͠ͳ͍͜ͱͱͳΔɻݪԻ੠Λ6400 HzҎԼʹଳҬ

੍ݶͯ͠΋Ի੠ͷݴޠ಺༰Λਖ਼͘͠ௌ͖ͱΔ͜ͱ͕ՄೳͰ͋Δ͜ͱ͸ࣄલʹ֬ೝ͍ͯ͠Δɻ·

ͨɺαϯϓϦϯάप೾਺16000 HzͰ࿥ԻՄೳͳप೾਺ൣғͷ্ݶ෇ۙͷप೾਺͸ɺͨͱ͑࿥

ԻͰ͖͍ͯͨͱͯ͠΋ຊདྷͱ͸ҟͳΔಛ௃ʹ࿪ΊΒΕ͍ͯΔ͓ͦΕ΋͋ΔɻҎ্ͷΑ͏ͳཧ༝

ʹΑΓɺ6400 Hz·ͰΛ෼ੳର৅ͷप೾਺ൣғͱܾΊͨɻ

·ͨɺ෼ੳʹྟքଳҬΛ༻͍Δ͜ͱͷੋඇʹ͍ͭͯड़΂͓ͯ͘ɻ ຊݚڀͷ໨త͸ɺԻ੠ͷ ύϫʔεϖΫτϧแབྷʹΈΒΕΔಛ௃ΛҼࢠͱͯ͠நग़͢Δ͜ͱͰ͋Δɻ෼ੳର৅ͱ͢ΔԻ੠

ͷجຊप೾਺͸150 Hzఔ౓Ͱ͋Δɻ௿Ҭʹ͓͍ͯ100 HzҎԼͷଳҬ෯ͱͳΔௌ֮ϑΟϧλ Λඞͣ͠΋༻͍Δඞཁ͸ͳ͍ɻྟքଳҬ෯Ͱ෼ׂ͞Εͨ20νϟϯωϧͷࡶԻۦಈԻ੠͕΄΅

׬શʹ໌ྎͰ͋ΔɻࡶԻۦಈԻ੠Λ߹੒͢Δ৔߹ʹ͓͍ͯ΋Ұͭͷνϟϯωϧ಺ͰεϖΫτϧ

͕ฏୱͰ͋Δํ͕؆ศͰ͋ΔɻҎ্ͷΑ͏ͳ͜ͱ͔ΒྟքଳҬΛٞ࿦ͷग़ൃ఺ʹ͢Δ͜ͱ͕ଥ

౰Ͱ͋Δͱߟ͑ΒΕΔͰ͋Ζ͏ɻ

(28)

ද 2.1 ྟքଳҬϑΟϧλͷத৺प೾਺ͱ௨աଳҬ Band no. Center frequency (Hz) Passband (Hz)

1 75 50–100

2 150 100–200

3 250 200–300

4 350 300–400

5 450 400–510

6 570 510–630

7 700 630–770

8 840 770–920

9 1000 920–1080

10 1170 1080–1270

11 1370 1270–1480

12 1600 1480–1720

13 1850 1720–2000

14 2150 2000–2320

15 2500 2320–2700

16 2900 2700–3150

17 3400 3150–3700

18 4000 3700–4400

19 4800 4400–5300

20 5800 5300–6400

ਤ 2.2ͷखଓ͖ͰಘΒΕͨ20ͷύϫʔมಈ͸20มྔ͔ΒͳΔଟมྔσʔλͱͯ͠ΈΔ͜ͱ

͕Ͱ͖Δɻ͜ͷଟมྔσʔλΛຊݚڀͰ৽͘͠ఏҊ͢Δʮى఺Ҡಈओ੒෼෼ੳʯʹ͔͚ɺओ੒

෼ΛऔΓग़͠ɺओ੒෼ͷҼࢠෛՙྔΛόϦϚοΫεճస(Kaiser, 1958)͢Δ͜ͱͰύϫʔεϖ ΫτϧҼࢠΛநग़ͨ͠ɻ

ຊདྷͷओ੒෼෼ੳ͸ɺଟ࣍ݩۭؒதʹදݱ͞Εͨଟมྔσʔλʹରͯ͠ɺͦͷσʔλͷ෼ࢄ

͕࠷େͱͳΔΑ͏ͳଟ࣍ݩ্ۭؒͷํ޲Λओ੒෼ͱͯ͠ॱ࣍ٻΊ͍ͯ͘෼ੳख๏Ͱ͋Δ(e.g.,

Jolliﬀe, 2002)ɻͦͷͨΊɺओ੒෼Λܾఆ͚ͮΔݻ༗ϕΫτϧ͸σʔλͷॏ৺Λى఺ͱͯ͠ٻΊ

ΒΕΔɻ͜ͷΑ͏ʹͯ͠ٻΊΒΕͨओ੒෼͕දݱ͠͏Δ৘ใ͚ͩͰݩͷଟมྔσʔλΛ࠶ߏ੒

͢Δͱ͍͏͜ͱ͸ɺݩͷଟมྔσʔλΛओ੒෼ۭؒʹਖ਼ࣹӨͨ͠΋ͷʹஔ͖׵͑Δͱ͍͏͜ͱ Ͱ͋Δɻ͜͜Ͱ΋͠ɺσʔλͷྵ఺ɺຊݚڀͷ৔߹͸ແԻΛද͢఺͕ओ੒෼෼ੳʹΑΔ෦෼ۭ

ؒʹؚ·Εͳ͔ͬͨ৔߹ɺແԻΛද͢఺͸σʔλΛ࠶ߏ੒͢Δࡍʹ࿪ΊΒΕɺྟքଳҬͷͲ͜

͔ʹύϫʔΛ࣋ͬͨ఺ʹҠΔɻ͜͏ͯ͠࠶ߏ੒͞ΕͨσʔλΛجʹ࠶߹੒ͨ͠Ի੠͸ఆৗతͳ

(29)

ࡶԻΛؚΉ͜ͱͱͳΔɻ͜ΕΛௌऔऀ͕ௌ͚͹ɺ࠶߹੒͞ΕͨԻ੠ͷதͰࡶԻ͕໐Γଓ͍͍ͯ

ΔΑ͏ʹײ͡ΔͰ͋Ζ͏ɻຊདྷҙਤ͠ͳ͍ఆৗతͳࡶԻ੒෼͕ੜ͡ΔԻ੠Λௌऔ࣮ݧʹ༻͍Δ ͷ͸ద੾Ͱ͸ͳ͍ɻ͓ͦΒ͘ɺಉ༷ͷఆৗࡶԻ͕ Zahorian and Rothenberg (1981)ͷ࠶߹੒

Ի੠ʹ͓͍ͯ΋ੜ͍ͯͨ͡ͱߟ͑ΒΕΔ͕ɺ͜ͷ͜ͱʹ͍ͭͯಛʹݴٴ͸͞Ε͍ͯͳ͔ͬͨɻ

͜Εʹରͯ͠ى఺Ҡಈओ੒෼෼ੳ͸ɺओ੒෼෼ੳʹΑͬͯٻΊΒΕΔ෦෼ۭؒΛఆٛ͢Δϕ Ϋτϧ͕σʔλͷॏ৺Ͱ͸ͳ͘ɺશͯͷมྔͷ஋͕ྵͱͳΔ఺ɺຊ࿦จͰ͸ແԻΛද͢఺Λى

఺ͱ͢ΔΑ͏ʹมܗͨ͠ख๏Ͱ͋Δ¹ɻ͜͏͢Δ͜ͱʹΑͬͯɺແԻΛද͢఺͸ɺσʔλͷ࠶

ߏ੒Λͯ͠΋ඞͣແԻͷ··ͱͳΓɺ্ड़ͷ໰୊ʹىҼ͢Δɺҙਤ͠ͳ͍ఆৗతͳࡶԻ੒෼͕

ൃੜ͢Δͱ͍͏͜ͱ͸ͳ͘ͳΔɻਤ 2.3͸ɺ̎มྔ͔ΒͳΔඇෛͷσʔλʹରͯ͠ɺ௨ৗͷओ

੒෼෼ੳͱى఺Ҡಈओ੒෼෼ੳΛͦΕͧΕߦͬͨ৔߹ͱͰɺࢉग़͞ΕΔओ੒෼͕ͲͷΑ͏ʹҟ ͳΔͷ͔Λࣔ֓͢೦ਤͰ͋Δɻ

V1 V2

Gravity center of data PC1

V1

V2 PC1

Conventional PCA Origin-shifted PCA

ਤ 2.3௨ৗͷओ੒෼෼ੳ(ࠨ)ͱى఺Ҡಈओ੒෼෼ੳ(ӈ)ʹ͓͚Δओ੒෼ࢉग़ͷ֓೦ਤɻ௨ৗ

ͷओ੒෼෼ੳͰ͸ɺଟ࣍ݩۭؒͷݪ఺ؚ͕·ΕΔΑ͏ʹओ੒෼͕ࢉग़͞Ε͓ͯΒͣɺݪ఺Λओ

੒෼࣠ʹਖ਼ࣹӨ͢ΔͱͣΕ͕ੜ͡Δɻ͜ͷͣΕ͕Ի੠ͷ࠶߹੒ʹ͓͍ͯ͸ఆৗࡶԻͷݪҼͱ ͳΔɻ

ى఺Ҡಈओ੒෼෼ੳͰಘΒΕͨओ੒෼ۭؒΛఆٛ͢ΔϕΫτϧɺ͢ͳΘͪҼࢠෛՙྔΛόϦ ϚοΫεճస͢Δ͜ͱͰɺى఺Ҡಈओ੒෼෼ੳʹΑͬͯಘΒΕͨύϫʔεϖΫτϧҼࢠΛΑΓ ղऍ͠΍͍͢ɺͭ·Γ֤Ҽࢠ͕ͲͷྟքଳҬͱؔ࿈͕ڧ͍ͷ͔͕෼͔Γ΍͍͢ܗʹ͢Δ͜ͱ

͕Ͱ͖ΔɻόϦϚοΫεճసΛୈԿओ੒෼·ͰؚΊͯߦ͏͔Λม͑Δ͜ͱͰɺ̕छྨͷύϫʔ εϖΫτϧҼࢠͷ૊Λಘͨɻྫ͑͹ɺ̏Ҽࢠ͔ΒͳΔύϫʔεϖΫτϧҼࢠΛಘ͍ͨ৔߹͸ɺ

1ओ੒෼ۭؒΛఆٛ͢ΔϕΫτϧͷى఺Λશͯͷมྔͷ஋͕ྵͱͳΔ఺ʹ͢ΔͨΊͷ࣮ࡍతͳํ๏͸͍͔ͭ͘

ߟ͑ΒΕΔɻຊ࿦จͰ͸ɺओ੒෼෼ੳʹ͔͚Δଟมྔσʔλʹූ߸Λٯసͤͨ͞ଟมྔσʔλΛͭͳ͛ͯɺσʔ λͷॏ৺Λมྔͷ஋͕ྵͱͳΔ఺ʹม͑Δͱ͍͏ํ๏Λ༻͍ͨɻ

(30)

ୈ̏ओ੒෼·ͰΛόϦϚοΫεճసͨ͠ɻ

2.2.3 ݁Ռͱߟ࡯

ਤ2.4ɺ2.5ɺ2.6͸̏ͭͷݴޠ͔ΒಘΒΕͨ̕૊ͷύϫʔεϖΫτϧҼࢠʹ͍ͭͯɺྟքଳ

Ҭ͝ͱͷҼࢠෛՙྔΛࣔͨ͠΋ͷͰ͋Δɻ·ͣ͸ɺಘΒΕͨύϫʔεϖΫτϧҼࢠͷಛ௃͕

̏ͭͷݴޠͷؒͰࣅ͍ͯΔ͔Ͳ͏͔Λݟ͍ͯ͘ɻ̐ҼࢠΛநग़ͨ͠ͱ͜Ζ·Ͱޓ͍ʹࣅͨҼࢠ ͷ഑ஔ͕ಘΒΕ͍ͯΔͷ͕෼͔Δɻ̑ҼࢠΛ௒͑ͯύϫʔεϖΫτϧҼࢠΛநग़͢Δͱɺݴ ޠؒͰڞ௨͍ͯ͠Δͱߟ͑ΒΕΔҼࢠΛݟ͚ͭΔͷ͕ࠔ೉ʹͳͬͨɻҎ্ͷ݁Ռ͸ Ueda and

Nakajima (2017)Ͱใࠂ͞Εͨ಺༰ͱҰக͍ͯ͠Δɻ

࣍ʹɺݴޠؒͰڞ௨ͨ͠ಛ௃ͷύϫʔεϖΫτϧҼࢠ͕ಘΒΕͨҼࢠ਺ͰɺͲͷΑ͏ͳಛ

௃Λ࣋ͬͨҼࢠ͕ಘΒΕͨͷ͔Λݟ͍ͯ͘ɻ̍Ҽࢠ෼ੳͰ͸ɺ͢΂ͯͷྟքଳҬʹ͓͍ͯҼࢠ

ෛՙྔ͕ਖ਼ͷ஋Ͱ͋ͬͨɻ͜Ε͸ى఺Ҡಈओ੒෼෼ੳʹΑͬͯಘΒΕΔୈ̍ओ੒෼(Ҽࢠ)͕ ඞͣ΋ͭಛ௃Ͱ͋Δɻ̎Ҽࢠ෼ੳͰ͸ɺ໿1000 HzΛத৺ͱ͢ΔதଳҬʹେ͖͍Ҽࢠෛՙྔ

Λ࣋ͭҼࢠͱͦͷ྆ଆͷଳҬͰҼࢠෛՙྔ͕େ͖͍Ҽࢠͱ͕ಘΒΕ͍ͯΔɻ ͞Βʹɺ̏Ҽࢠ

෼ੳͰ͸ɺ໿1000 HzΛத৺ͱ͢ΔதଳҬʹେ͖͍ҼࢠෛՙྔΛ࣋ͭҼࢠɺ໿3000 HzҎ্

ͷߴଳҬʹେ͖͍ҼࢠෛՙྔΛ΋ͭҼࢠɺͦͯ͠໿500 HzҎԼͷ௿Ҭ͓ΑͼɺதଳҬͱߴଳ ҬͷؒͷଳҬ(໿1500–3000 Hz)ʹ෼͔Εͯେ͖͍ҼࢠෛՙྔΛ࣋ͭೋๆੑͷҼࢠ͕ͦΕͧΕ ಘΒΕ͍ͯΔɻೋๆੑͷҼࢠ͕ݱΕΔͱ͍͏ಛ௃͸ɺUeda and Nakajima (2017)ͷ෼ੳͰ΋

ಉ༷ʹใࠂ͞Ε͍ͯΔɻຊ෼ੳ͕ຊ࣭తʹ͸Ueda and Nakajima (2017)ͷ෼ੳͱಉ݁͡ՌΛ ಋ͍͍ͯΔ͜ͱΛࣔ͢ࢦඪͰ͋Δͱݴ͑ΔͰ͋Ζ͏ɻͦͯ̐͠Ҽࢠ෼ੳͰ͸ɺ̏Ҽࢠ෼ੳͷͱ

͖ʹಘΒΕͨதଳҬʹେ͖͍ҼࢠෛՙྔΛ࣋ͭҼࢠͱߴଳҬʹେ͖͍ҼࢠෛՙྔΛ࣋ͭҼࢠ ʹՃ͑ɺೋๆੑͷҼࢠ͕ೋͭͷҼࢠʹ෼͔ΕͨΑ͏ͳҼࢠ͕ಘΒΕ͍ͯΔɻ͜ͷ݁Ռ΋·ͨɺ

Ueda and Nakajima (2017)ͷ෼ੳ݁ՌͱҰக͍ͯ͠Δɻ͕ͨͬͯ͠ɺࠓճಋೖͨ͠ى఺Ҡಈ

ओ੒෼෼ੳʹΑͬͯ΋ɺઌߦݚڀͱಉ౳ͷύϫʔεϖΫτϧҼࢠ͕ಘΒΕͨͱ൑அͰ͖Δɻ

͋ΔྟքଳҬʹ͓͍ͯɺಛఆͷҼࢠͷҼࢠෛՙྔͷઈର஋͕େ͖͘ɺͦΕʹൺ΂ͯͦΕҎ֎

ͷҼࢠͷҼࢠෛՙྔͷઈର஋͕খ͍͞৔߹͸ɺͦͷଳҬͷύϫʔͷมԽ͕Ҽࢠෛՙྔͷઈର஋

͕େ͖ͳҼࢠʹΑͬͯઆ໌͞ΕΔׂ߹͕େ͖͍͜ͱΛҙຯ͢Δɻ·ͨผͷྟքଳҬʹ͓͍ͯ΋

ಉ͡Α͏ͳҼࢠෛՙྔͷؔ܎ʹͳ͍ͬͯΔ৔߹ɺͦΕΒͷෳ਺ͷଳҬʹ͓͍ͯύϫʔ͕ಉ͡Α

͏ʹมԽ͢Δ͜ͱͱͳΔɻྫ͑͹̏Ҽࢠ෼ੳͷ1000 Hz෇ۙͷ͍͔ͭ͘ͷྟքଳҬʹ͓͍ͯ

͸ɺҰͭͷҼࢠ(നൈ͖ͷؙͰදͨ͠΋ͷ)ͷҼࢠෛՙྔ͕ਖ਼ͷ஋Ͱେ͖͘ɺͦΕҎ֎ͷҼࢠ

(31)

Center frequency of critical bands (Hz)

1-factor2-factor3-factor

Factor loading

British English Japanese Mandarin Chinese

ਤ 2.4ى఺Ҡಈओ੒෼෼ੳʹج͍ͮͯಘΒΕͨύϫʔεϖΫτϧҼࢠͷɺྟքଳҬ͝ͱͷҼ ࢠෛՙྔɻԣ࣠ͷ਺஋͸50–6400 Hzͷൣғͷप೾਺ʹ഑ஔ͞Εͨ20ͷྟքଳҬͷ֤த৺प೾

਺Λ͍ࣔͯ͠ΔɻྟքଳҬͷத৺प೾਺͓ΑͼͦͷଳҬ෯ʹ͍ͭͯ͸ɺද 2.1Λࢀরͷ͜ͱɻ ࠨ͔ΒɺΠΪϦεӳޠ฼ޠ࿩ऀɺ೔ຊޠ฼ޠ࿩ऀɺதࠃޠ(ී௨࿩)฼ޠ࿩ऀͷ݁ՌΛࣔ͢ɻ্

ஈ͔Βɺ1Ҽࢠɺ2Ҽࢠɺ3ҼࢠɺΛͦΕͧΕநग़ͨ͠৔߹ͷ݁ՌͰ͋Δɻ

(32)

Factor loading

(33)

Factor loading

(34)

ͷҼࢠෛՙྔ͸ྵʹ͍ۙͱ͍͏ಛ௃͕ڞ௨͍ͯ͠Δ(ਤ 2.4)ɻΑͬͯ͜ΕΒͷଳҬͷύϫʔ͸

·ͱ·ͬͯมಈ͍ͯ͠Δͱղऍ͢Δ͜ͱ͕Ͱ͖Δɻ

͞Βʹ̐Ҽࢠ෼ੳʹ͓͍ͯɺ໿500 HzҎԼͷ̑ͭͷྟքଳҬͰҼࢠෛՙྔͷେ͖͍Ҽࢠ͸ɺ

ͦΕΒͷ̑ͭͷଳҬʹ͓͍ͯಉఔ౓ͷҼࢠෛՙྔͰ͋ͬͨɻέϓετϥϜ෼ੳʹΑͬͯύϫʔ εϖΫτϧͷԻݯಛੑ͕ద੾ʹऔΓআ͔ΕͨͨΊͰ͋Ζ͏ɻ

෼ੳ͢Δଟมྔσʔλͷߏ଄(ଟ࣍ݩ্ۭؒͰͷσʔλͷ෼෍)ʹΑͬͯ͸ɺ௨ৗͷओ੒෼

෼ੳͱى఺Ҡಈओ੒෼෼ੳͱͰ͸େ͖͘ҟͳΔҼࢠΛಋ͘Մೳੑ͕͋Δɻ̏Ҽࢠ෼ੳͱ̐Ҽࢠ

෼ੳͰಘΒΕͨύϫʔεϖΫτϧҼࢠ͕ɺ௨ৗͷओ੒෼෼ੳʹج͍ͮͯಘΒΕͨରԠ͢Δύ ϫʔεϖΫτϧҼࢠͱࣅͨΑ͏ಛ௃Λ͍࣋ͬͯΔͱ͍͏͜ͱ͸ɺྟքଳҬ͝ͱͷύϫʔεϖΫ τϧͷมಈͷσʔλΛଟ࣍ݩ্ۭؒʹදݱͨ͠ࡍʹɺσʔλͷॏ৺ͱແԻ఺(ଟ࣍ݩۭؒͷݪ

఺)ͱΛ݁Ϳ௚ઢ্෇ۙʹσʔλ͕෼෍͍ͯͨ͠ͱ͍͏͜ͱʹͳΔɻ

ද2.2ʹى఺Ҡಈओ੒෼෼ੳʹج͍ͮͯಘΒΕͨҼࢠɺ͓Αͼ௨ৗͷओ੒෼෼ੳʹج͍ͮͯ

ಘΒΕͨҼࢠͷྦྷੵد༩཰Λࣔ͢ɻد༩཰ͱ͸ɺ͋Δओ੒෼·ͨ͸Ҽࢠ͕ݩͷଟมྔσʔλ ͷ෼ࢄΛͲΕ͚ͩͷׂ߹Ͱอ͍࣋ͯ͠Δͷ͔Λࣔ͢΋ͷͰ͋ΔɻͦΕͧΕͷओ੒෼·ͨ͸Ҽ ࢠ͕ɺݩͷଟมྔσʔλͷ৘ใΛͲΕ͚ͩઆ໌͍ͯ͠Δ͔Λද͢΋ͷͱ΋ߟ͑Δ͜ͱ͕Ͱ͖

Δɻྦྷੵد༩཰͸ͦͷد༩཰Λ͋Δओ੒෼਺ɺҼࢠ਺·ͰͰྦྷੵͨ͠΋ͷͰ͋Δɻى఺Ҡಈओ

੒෼෼ੳʹج͍ͮͯಘΒΕͨୈ̍Ҽࢠͷد༩཰͸17–22%ఔ౓Ͱ͋ΓɺҼࢠ਺͕૿Ճ͢Δʹ͠

͕ͨͬͯྦྷੵد༩཰͸؇΍͔ʹ্ঢ͠ɺୈ̕Ҽࢠ·ͰͰ74–77%ఔ౓·Ͱ্ঢͨ͠ɻ֤ݴޠͰ ύϫʔεϖΫτϧҼࢠͷಛ௃͕ڞ௨͍ͯ͠Δୈ̐Ҽࢠ·Ͱͷྦྷੵد༩཰͸49–56%ఔ౓Ͱ͋ͬ

ͨɻԻ੠ͷύϫʔεϖΫτϧมಈͷಛ௃ͷ͓Αͦ൒෼͕ୈ̐Ҽࢠ·ͰͰઆ໌Ͱ͖ɺ͞Βʹ͸̏

ͭͷݴޠؒͰͦͷಛ௃͕ڞ௨͍ͯ͠Δͱ͍͏͜ͱΛ͍ࣔͯ͠Δɻ

ݴޠ͝ͱʹɺ௨ৗͷओ੒෼෼ੳͱى఺Ҡಈओ੒෼෼ੳͷྦྷੵد༩཰Λಉ͡Ҽࢠ਺ͷؒͰൺֱ

͢Δͱɺͦͷࠩ͸̎%ҎԼͰ͋ͬͨɻྦྷੵد༩཰ͱ͍͏؍఺Ͱ΋ɺى఺Ҡಈओ੒෼෼ੳʹجͮ

͍ͯಘΒΕͨύϫʔεϖΫτϧҼࢠ͕௨ৗͷओ੒෼෼ੳʹج͍ͮͯಘΒΕͨύϫʔεϖΫτ ϧҼࢠͱಉ౳Ͱ͋Δͱݴ͑ΔͩΖ͏ɻ

࿈ଓతʹൃ࿩ͨ͠Ի੠ͷύϫʔεϖΫτϧมಈͷσʔλ͸ɺى఺Ҡಈओ੒෼෼ੳΛ༻͍Δͷ ʹద͍ͯ͠Δͱݴ͑ΔɻҰํͰɺ฼Իͷఆৗ෦ͷύϫʔεϖΫτϧΛ฼Ի͝ͱʹҰͭҰͭ؍ଌ

͠ɺͦΕΒͷύϫʔεϖΫτϧͷप೾਺ଳҬ͝ͱͷϨϕϧΛมྔʹ༻͍Δ৔߹͸ɺى఺Ҡಈओ

੒෼෼ੳ͸ద͞ͳ͍ͱߟ͑ΒΕΔɻͪΐ͏ͲPlomp et al. (1967)͕෼ੳର৅ͱͨ͠Ի੠͕ͦΕ ʹ͋ͨΔɻ฼Իͷఆৗ෦͸ύϫʔͷมԽ͕҆ఆ͓ͯ͠Γɺ฼ԻΛҰԻͣͭൃ࿩ͨ͠৔߹ʹ͸ɺ

(35)

࿩ऀ͕׶͑ͯͦ͏͠Α͏ͱ͠ͳ͍ݶΓɺ฼Ի͝ͱʹύϫʔ͕େ͖͘ҟͳΔͱ͍͏͜ͱ͸ͳ͍Ͱ

͋Ζ͏ɻΑͬͯ؍ଌσʔλͷଟ࣍ݩ্ۭؒͷ෼෍͸σʔλͷॏ৺ͱۭؒͷݪ఺ͱΛ݁Ϳ௚ઢ্

෇ۙʹ͸෼෍͠ͳ͍ɻ͜ͷΑ͏ͳσʔλʹରͯ͠௨ৗͷओ੒෼෼ੳΛߦ͑͹ɺୈ̍ओ੒෼ͷҼ ࢠෛՙྔͷ͍͔͕ͭ͘ෛͷ஋²ͱͳΔ͜ͱ͕༧ଌ͞ΕΔ͕ɺಉ͡σʔλʹى఺Ҡಈओ੒෼෼ੳ Λߦ͑͹ɺୈ̍ओ੒෼ͷҼࢠෛՙྔ͸͢΂ͯਖ਼ͷ஋ͱͳΓɺ෼ੳ݁Ռ͕େ͖͘ҟͳΔ͓ͦΕ͕

͋Δɻ

௨ৗͷओ੒෼෼ੳʹΑͬͯಘΒΕͨύϫʔεϖΫτϧҼࢠ͔ΒԻ੠Λ࠶߹੒ͨ͠৔߹ʹੜ

͡ΔఆৗࡶԻͷྫΛਤ 2.7ʹݟΔ͜ͱ͕Ͱ͖ΔɻݪԻ੠͕ࡶԻۦಈԻ੠ͱͯ͠࠶߹੒͞Εͨ΋

ͷͰ͋Δɻ࣮ࡍͷ࠶߹੒ͷํ๏ʹ͍ͭͯ͸࣍અͰৄ͘͠ड़΂Δɻ͜ͷਤͰࣔ͢ྫͷݪԻ੠͸

1.9–2.0 s෇͕ۙ΄΅ແԻঢ়ଶͰ͋Δɻ͜ΕΛ௨ৗͷओ੒෼෼ੳͰಘΒΕͨ̐Ҽࢠ͔Β࠶߹੒

ͨ͠৔߹ɺ͓Αͦ1000–1500 HzͷଳҬʹఆৗతͳࡶԻ͕ੜ͍ͯ͡Δͷ͕෼͔ΔɻҰํɺى఺

Ҡಈओ੒෼෼ੳͰಘΒΕͨҼࢠ͔Β࠶߹੒ͨ͠৔߹͸ɺͦͷΑ͏ͳఆৗతͳࡶԻ͸ੜ͍ͯ͡

ͳ͍ɻ

2Ҽࢠෛՙྔͷූ߸ࣗମʹ͸ҙຯ͸ͳ͍ɻ͢΂ͯͷූ߸ΛೖΕସ͑ͯ΋ओ੒෼·ͨ͸Ҽࢠ͕ද͢΋ͷ͸ಉ͡Ͱ

͋Δɻූ߸ͷҧ͍͕ҙຯΛ΋ͭͷ͸ɺҰͭͷओ੒෼·ͨ͸Ҽࢠ಺ʹ͓͍ͯɺมྔؒͰҼࢠෛՙྔΛൺֱ͢Δ৔߹

Ͱ͋ΔɻΑͬͯɺى఺Ҡಈओ੒෼෼ੳͰಘΒΕΔୈ̍ओ੒෼ͷҼࢠෛՙྔ͸͢΂ͯෛͷ஋ͱͳΔͱݴ͍׵͑Δ͜

ͱ΋Ͱ͖Δɻຊ࿦จͰ͸෼ੳ݁ՌΛݟ΍͘͢͢ΔͨΊʹɺͦΕͧΕͷҼࢠʹ͍ͭͯɺҼࢠෛՙྔͷઈର஋͕࠷΋

େ͖͍΋ͷ͕ਖ਼ͷ஋ͱͳΔΑ͏ʹූ߸ΛͦΖ͑ͯදࣔͨ͠ɻ

(36)

ද 2.2 ى఺Ҡಈओ੒෼෼ੳͱ௨ৗͷओ੒෼෼ੳͰಘͨҼࢠͷྦྷੵد༩཰ɻ Japanese

Cumulative contribution (%) Number of factors Conventional Proposal

1 23.3 22.9

2 38.0 37.3

3 49.9 47.9

4 55.9 55.6

5 61.7 61.3

6 66.5 66.3

7 70.7 70.5

8 74.3 74.1

9 77.7 77.6

British English

1 22.4 21.5

2 35.3 34.5

3 48.7 45.6

4 54.0 53.7

5 59.9 59.7

6 65.2 64.5

7 69.2 68.8

8 73.3 72.3

9 76.5 76.3

Mandarin Chinese

1 17.9 17.3

2 32.0 31.5

3 43.1 40.7

4 48.8 48.6

5 55.7 55.3

6 61.2 60.3

7 65.8 65.4

8 70.3 69.3

9 73.8 73.7

日本語音声におけるパワースペクトル因子の音声知 覚上の役割

九州大学学術情報リポジトリ

Kyushu University Institutional Repository