The trough impact: Can we foresee tongue bringing down from acoustic information alone .


27 views
Uploaded on:
Category: General / Misc
Description
The trough effect: Can we predict tongue lowering from acoustic data alone?. Yolanda Vazquez Alvarez. Overview. 1. Background on the ‘trough effect’ 2. Aim of this experiment 3. Experimental method & results 4. Acoustic-to-articulatory mapping 5. Conclusions.
Transcripts
Slide 1

The trough impact: Can we anticipate tongue bringing down from acoustic information alone? Yolanda Vazquez Alvarez

Slide 2

Overview 1. Foundation on the \'trough impact\' 2. Point of this test 3. Trial strategy & comes about 4. Acoustic-to-articulatory mapping 5. Conclusions

Slide 3

Background – The \'trough impact\' The "trough" impact happens in symmetrical VCV groupings and has been depicted as: \'A Momentary deactivation of the tongue development amid the consonant conclusion\' (Bell-Berti, F. & Harris, K.; Gay, T)

Slide 4

Background – Acoustic proof Lindblom et al. (2002) gathered direct measures of the F2 directions from symmetrical VCV articulations (V=/i/)

Slide 5

Background – Ultrasound confirm Used QMUC\'s information from the trough test 3 Annotation guides relating toward 3 diverse tongue forms. 2 Measurements of tongue removal (MTD) were completed for these 3 unique forms

Slide 6

Background – Ultrasound prove MTDs were essentially not quite the same as each other for/iCi/arrangements ( ipi (t (9) = - 8.295, p< 0.010), ibi (t (9) = - 9.774, p< 0.010)

Slide 7

Background – Advantages & inconveniences of both methods Acoustics: -Good time determination -Doesn\'t require particular hardware to obtain the information No perception of the tongue Ultrasound: -Tongue shape perception -Physical estimation Need for edge by-edge examination of the tongue recording

Slide 8

Aim of this trial Given the benefits of acoustic estimations: How certain would we be able to be that the acoustic estimation of the tongue bringing down gives us a genuine representation of the trough impact?

Slide 9

Experimental strategy Subjects 5 local speakers of English, different accents Data symmetrical VCV groupings: C=/p/,/b/& V=/i/Repetitions 2 reps, n=20 Acoustic examination 4 F2 comment focuses: V 1 mid, V 1 counterbalance and F2 onset, V 2 mid 2 F2 estimations: F2V 1 - C and F2C-V 2 Ultrasound investigation 3 explanation focuses: V 1 mid, Cmid and V 2 sym 2 separate estimations: V 1 - C and C-V 2

Slide 10

Experiment - Results Correlation of F2 qualities and ultrasound information (V 1 - C) Pearson connection of V 1 - C and F2V 1 - C was huge (r (18)= .496, r2= 0.25, p<.05), foreseeing 25% of tongue bringing down difference. Utilizing both F2 indicators demonstrated an expansion in the relationship coefficient for V 1 - C, anticipating a 43% of tongue bringing down change. Pearson relationship of C-V 2 and F2C-V 2 was not huge.

Slide 11

Experiment – Results 3 conceivable reasons why we couldn\'t foresee the ascent for C-V 2 : Start of the tongue rise is in the conclusion so F2 can\'t indicate data about its conceivable development The measuring point was mostly on the discharge for/p/in the ultrasound information however we utilized V 2 mid in light of the fact that else we wouldn\'t have adequate F2 information Ultrasound time determination might be excessively poor, making it impossible to catch the ascending of the tongue at the suitable minute

Slide 12

Acoustic-to-articulatory mapping Korin Richmond et al. (2003) at CSTR, Edinburgh Univ., utilized a multilayer perceptron (MLP) neural system to assess articulatory directions The neural system was prepared on articulatory information (EMA) and acoustic information where articulatory component vectors (x,y) were standardized to lie in the range [0.1,0.9]

Slide 13

Acoustic-to-articulatory mapping The MLP was connected to the acoustic information from the ultrasound explore Despite being prepared on an alternate speaker, the trough marvels could be seen in the MLP gauges for the y-directions of tongue body development

Slide 14

MLP plot for/ibi/v 1 mid V 2 sym Cmid MLP plot for/ipi/Acoustic-to-articulatory mapping Annotation times from the ultrasound estimation focuses were utilized to analyze the evaluated tongue positions from the MLP A tongue bringing down and rising was seen in the MLP plots yet no significant factual results were gotten

Slide 15

Conclusions Acoustic data (F2) might miss for urgent articulatory development. It is difficult to guide acoustic change into articulatory change Current ultrasound time determination can be excessively poor, making it impossible to give data of quick articulatory change However, a joined approach can enhance both methods

Slide 16

References Bell-Berti, F. & Harris, K. (1974). More on the engine association of discourse motions. Haskins Laboratories: Status Report on Speech Research SR-37/38, 73-77. Gay, T. (1975). Some electromyographic measures of coarticulation in VCV-articulations. Haskins Laboratories: Status Report on Speech Research SR-44, 137-145. Lindblom, B., Sussman, H., Modarressi, G. & Burlingame, E. (2002). The trough impact: Implications for engine programming, Phonetica , 59, 245-262. K. Richmond, S. King, and P. Taylor. (2003). Displaying the vulnerability in recuperating enunciation from acoustics. PC Speech and Language , 17:153-172.

Slide 17

Acknowledgments Thanks go to SHS at QMUC in Edinburgh for the utilization of the ultrasound information from the trough explore. Additionally, I might want to say thanks to Korin Richmond at CSTR in Edinburgh for his advantage and help with the preparing of the acoustic information utilizing the MLP neural system.

Recommended
View more...