Proceedings Fonetik 2009 - Institutionen för lingvistik

More documents

Recommendations

Info

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm UniversityRecognition/classification40%30%20%10%0%-10%-20%-30%-40%1 3 5 7 9 11 13 15 17 19 21the two types of animations correctly, withμ=48% at chance level. The picture is to someextent altered when considering the discriminationscore, ∆=0.66 (standard deviation 0.12) forthe group. Fig. 5 shows that the variation betweensubjects is large. Whereas about half ofthem were close to chance level, the other halfwere more proficient in the discrimination task.The classification score was slightly influencedby the audio signal, since the subjectsclassified synthetic movements accompaniedby vocoded audio 6.5% more correctly than ifthey were accompanied by normal audio. Therewas no difference for the real movements. Itshould be noted that it was not the case that thesubjects consciously linked normal audio to themovements they believed were real, i.e., subjectswith low μ (and high ∆) did not differfrom subjects with high or chance-level μ.Fig. 5 also illustrates that the relation betweenthe classification score for individualsubjects and their difference in AVR-AVSword recognition in the perception test is weak.Subject 1 was indeed more aware than the averagesubject of what real tongue movementslook like and subjects 8, 10 and 11 (subjects 15,16 and 18 in Fig. 3), who had a negativeweighted AVR-AVS difference in word recognition,were the least aware of the differencesbetween the two conditions. On the other hand,for the remaining subjects there is very littlecorrelation between recognition and classificationscores. For example, subjects 4, 5 and 9were much more proficient than subject 1 atdiscriminating between the two animationtypes, and subject 2 was no better than subject7, even though there were large differences inperception results between them.δμAVR-AVSFigure 5. Classification score δμ relativechance-level (δμ=μ-0.5). The x-axis crosses atchance level and the bars indicate scores aboveor below chance. For subjects 1–11, who participatedin the perception test, the weighted differencein word recognition rate between theAVR and AVS conditions is also given.DiscussionThe perception test results showed that animationsof the intraoral articulation may be validas a speech perception support, since the wordrecognition score was significantly higher withanimations than without. We have in this testnot investigated if it is specifically the displayof tongue movements that is beneficial. Theresults from Wik & Engwall (2008) and Badinet al. (2008) suggest that a normal view of theface without any tongue movements visiblewould be as good or better as a speech perceptionsupport. The results of the current studyhowever indicate that animations based on realmovements were significantly higher, and weare therefore currently working on a new coarticulationmodel for the tongue, based on EMAdata, in order to be able to create sufficientlyrealistic synthetic movements, with the aim ofproviding the same level of support as animationsfrom real measurements.The classification test results suggest thatsubjects are mostly unaware of what real tonguemovements look like, with a classificationscore at chance level. They could to a largerextent discriminate between the two types ofanimations, but still at a modest level (2/3 ofthe animations correctly separated).In the explanations of what they had lookedat to judge the realism of the tongue movements,two of the most successful subjectsstated that they had used the tongue tip contactwith the palate to determine if the animationwas real or not. However, subjects who had lowμ, but high ∆ or were close to chance level alsostated that they had used this criterion, and itwas hence not a truly successful method.An observation that did seem to be useful todiscern the two types of movements (correctlyor incorrectly labeled) was the range of articulation,since the synthetic movements were larger,and, as one subject stated, “reached theplaces of articulation better”. The subject withthe highest classification rate and the two withthe lowest all used this criterion.A criterion that was not useful, ventured byseveral subjects who were close to chance, wasthe smoothness of the movement and the assumptionthat rapid jerks occurred only in thesynthetic animations. This misconception israther common, due to the rapidity and unfamiliarityof tongue movements: viewers arevery often surprised by how fast and rapidlychanging tongue movements are.34
Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm UniversityConclusionsThe word recognition test of sentences with degradedaudio showed that animations based onreal movements resulted in significantly betterspeech perception than rule-based. The classificationtest then showed that subjects were unableto tell if the displayed animated movementswere real or synthetic, and could to amodest extent discriminate between the two.This study is small and has several factorsof uncertainty (e.g., variation between subjectsin both tests, the influence of the face movements,differences in articulatory range of thereal and rule-based movements) and it is hencenot possible to draw any general conclusions onaudiovisual speech perception with augmentedreality. It nevertheless points out a very interestingpath of future research: the fact that subjectswere unable to tell if animations were createdfrom real speech movements or not, butreceived more support from this type of animationsthan from realistic synthetic movements,gives an indication of a subconscious influenceof visual gestures on speech perception. Thisstudy cannot prove that there is a direct mappingbetween audiovisual speech perceptionand speech motor planning, but it does hint atthe possibility that audiovisual speech is perceivedin the listener’s brain terms of vocaltract configurations (Fowler, 2008). Additionalinvestigations with this type of studies couldhelp determine the plausibility of differentspeech perception theories linked to the listener’sarticulations.AcknowledgementsThis work is supported by the Swedish ResearchCouncil project 80449001 Computer-Animated LAnguage TEAchers (CALATEA).The estimation of parameter values from motioncapture and articulography data was performedby Jonas Beskow.ReferencesAgelfors, E., Beskow, J., Dahlquist, M., Granström,B., Lundeberg, M., Spens, K.-E. andÖhman, T. (1998). Synthetic faces as alipreading support. Proceedings of ICSLP,3047–3050.Badin, P., Tarabalka, Y., Elisei, F. and Bailly,G. (2008). Can you ”read tongue movements”?,Proceedings of Interspeech, 2635–2638.Benoît, C. and LeGoff, B. (1998). Audio-visualspeech synthesis from French text: Eightyears of models, design and evaluation atthe ICP. Speech Communication 26, 117–129.Beskow, J. (1995). Rule-based visual speechsynthesis. Proceedings of Eurospeech, 299–302.Beskow, J., Engwall, O. and Granström, B.(2003). Resynthesis of facial and intraoralmotionfrom simultaneous measurements.Proceedings of ICPhS, 431–434.Branderud, P. (1985). Movetrack – a movementtracking system, Proceedings of the French-Swedish Symposium on Speech, 113–122,Engwall, O. (2003). Combining MRI, EMA &EPG in a three-dimensional tongue model.Speech Communication 41/2-3, 303–329.Fowler, C. (2008). The FLMP STMPed, PsychonomicBulletin & Review 15, 458–462.Grauwinkel, K., Dewitt, B. and Fagel, S.(2007). Visual information and redundancyconveyed by internal articulator dynamicsin synthetic audiovisual speech. Proceedingsof Interspeech, 706–709.Liberman A, Cooper F, Shankweiler D andStuddert-Kennedy M (1967). Perceptionof the speech code. Psychological Review,74, 431–461.Liberman A & Mattingly I (1985). The motortheory of speech perception revised.Cognition, 21, 1–36.Siciliano, C., Williams, G., Beskow, J. andFaulkner, A. (2003). Evaluation of a multilingualsynthetic talking face as a communicationaid for the hearing impaired, Proceedingsof ICPhS, 131–134.Skipper J., Wassenhove V. van, NusbaumH. and Small, S. (2007). Hearing lipsand seeing voices: how cortical areassupporting speech production mediateaudiovisual speech perception. CerebralCortex 17, 387 – 2399.Sumby, W. and Pollack, I. (1954). Visual contributionto speech intelligibility in noise,Journal of the Acoustical Society of America26, 212–215.Traunmüller, H. (2007). Demodulation, mirrorneurons and audiovisual perception nullifythe motor theory. Proceedings of Fonetik2007, KTH-TMH-QPSR 50: 17–20.Wik, P. and Engwall, O. (2008). Can visualizationof internal articulators support speechperception?, Proceedings of Interspeech2008, 2627–2630.35
Page 1 and 2: Department of LinguisticsProceeding
Page 3 and 4: Proceedings, FONETIK 2009, Dept. of
Page 33: Proceedings, FONETIK 2009, Dept. of
Page 41 and 42: Proceedings, FOETIK 2009, Dept. of
Page 85 and 86:
Proceedings, FONETIK 2009, Dept. of
Page 87 and 88:
Page 89 and 90:
Page 91 and 92:
Page 93 and 94:
Page 95 and 96:
Page 97 and 98:
Page 99 and 100:
Page 101 and 102:
Page 103 and 104:
Page 105 and 106:
Page 107 and 108:
Page 109 and 110:
Page 111 and 112:
Page 113 and 114:
Page 115 and 116:
Page 117 and 118:
Page 119 and 120:
Page 121 and 122:
Page 123 and 124:
Page 125 and 126:
Page 127 and 128:
Proceedings, FOETIK 2009, Dept. of
Page 129 and 130:
Page 131 and 132:
Page 133 and 134:
Page 135 and 136:
Page 137 and 138:
Page 139 and 140:
Page 141 and 142:
Page 143 and 144:
Page 145 and 146:
Page 147 and 148:
Page 149 and 150:
Page 151 and 152:
Page 153 and 154:
Page 155 and 156:
Page 157 and 158:
Page 159 and 160:
Page 161 and 162:
Page 163 and 164:
Page 165 and 166:
Page 167 and 168:
Page 169 and 170:
Page 171 and 172:
Page 173 and 174:
Page 175 and 176:
Page 177 and 178:
Page 179 and 180:
Page 181 and 182:
Page 183 and 184:
Page 185 and 186:
Page 187 and 188:
Page 189 and 190:
Page 191 and 192:
Page 193 and 194:
Page 195 and 196:
Page 197 and 198:
Page 199 and 200:
Page 201 and 202:
Page 203 and 204:
Page 205 and 206:
Page 207 and 208:
Page 209 and 210:
Page 211 and 212:
Page 213 and 214:
Page 215 and 216:
Page 217 and 218:
Page 219 and 220:
Page 221 and 222:
Page 223 and 224:
Page 225 and 226:
Page 227:
Department of LinguisticsPhonetics
show all

Proceedings Fonetik 2009 - Institutionen för lingvistik

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?