On the relationship between input parameters in the two-mass - PowerPoint PPT Presentation

On the relationship between input parameters in the two-mass vocal-fold model with acoustical coupling and signal parameters of the glottal flow A. Van Hirtum 1 , I. Lopez 2 , M. Hirschberg 2 , X. Pelorson 1 1 ICP, INPG, Grenoble, France 2 TUE, Eindhoven, The Netherlands

Introduction voiced sound synthesis production analysis generative model physical model glottal volume air flow (Ug)

Introduction generative model physical model acoustical signal physiology & fluid flow simplification of reality generative function two-mass model glottal volume air flow (Ug)

Objective To assess the sensitivity of the two-mass model with acoustical coupling to the input-parameters

Objective input parameters (I) signal parameters (S) (physiology, fluid flow, (acoustical signal, intrinsic) intrinsic) physical two-mass generative model (Lous) function (LF) glottal volume flow Ug glottal volume flow Ug voice quality parameters (time)

Two-mass vocal fold model vocal fold sub-glottal supra-glottal h c (trachea h 0 ) (vocal tract h 3 ) y K 1 , ξ 1 Psub m tot K 2 , ξ 2 x physiology mechanics fluid flow

Vocal tract configuration /a/ /i/

Signal time-parameters /a/ (S) U 0 , D 0 , F 0 , T e , T a , A g (h min ) Q 0 =t 0 /T 0 Q s =t p /t n Q r =A p /A n Q m =U 0 /D 0

Relative sensitivity index − (S S )/S = var ref ref β [-] − (I I )/I var ref ref

Two-mass model input parameter reference values (I) L w 10 mm L g 3 mm 0.0002 L in , L out h 0 12 mm mm m 1 , m 2 0.1 g m tot 0.2 g K 1 , K 2 40 N/m K c 20 N/m ξ 1 , ξ 2 0.1 ξ c 1.1 y 1 , y 2 0.01 mm h 3 /a/, /i/ P sub 1000 Pa x 1 /x 2 0.06

Reference signal parameters from Ug (dUg) F 0 Q 0 Q s Q r U 0 D 0 Q m T e T a h min [Hz] [-] [-] [-] [m 3 /s] [m 3 /s 2 ] [s] [s] [s] [m] /a/ 108 0.69 1.78 1.37 0.0002 0.24 0.0022 0.0010 0.014 0.0006 /i/ 104 0.70 3.18 2.90 0.0002 0.29 0.0015 0.0007 0.0047 0.0005

Sensitivity /i/ F 0 Q 0 /a/ Q s Q r U 0 D 0 Q m T e T a h min >600

/i/ F 0 Q 0 /a/ Q s Q r U 0 D 0 Q m T e T a h min /i/ ≠ /a/

/a/ F 0 Q 0 /i/ Q s Q r U 0 D 0 Q m T e T a h min

/a/ F 0 Q 0 /i/ Q s Q r U 0 D 0 Q m T e T a h mi > β (K) β (K ) c

Conclusion • sensitivity of the input-parameters of physical two-mass model on signal-parameters is quantified in terms of vocal quality parameters and 4 Liljencrants-Fant input parameters • influence of vocal tract shape on the sensitivity

Introduction voiced sound synthesis production analysis glottal volume air flow (Ug)

Introduction glottal volume air flow (Ug) synthesis production analysis generative model physical model