Paper Introduction. ~ Modelling the Uncertainty in Recovering Articulation from Acoustics ~ Korin Richmond, Simon King, and Paul Taylor.

Similar documents
10.3 Autocorrelation Function of Ergodic RP 10.4 Power Spectral Density of Ergodic RP 10.5 Normal RP (Gaussian RP)

Intro to multivariate AR(1) models estimating interaction strengths, aka the B matrix

Processamento Digital de Sinal

An Analysis of State Evolution for Approximate Message Passing with Side Information

Discrete-Time Signals and Systems. Introduction to Digital Signal Processing. Independent Variable. What is a Signal? What is a System?

Exercise 3 Stochastic Models of Manufacturing Systems 4T400, 6 May

Ideal Amplifier/Attenuator. Memoryless. where k is some real constant. Integrator. System with memory

Introduction to Mobile Robotics Mapping with Known Poses

b : the two eigenvectors of the Grover iteration Quantum counting algorithm

Comparisons Between RV, ARV and WRV

A Two-Level Quantum Analysis of ERP Data for Mock-Interrogation Trials. Michael Schillaci Jennifer Vendemia Robert Buzan Eric Green

WILD Estimating Abundance for Closed Populations with Mark-Recapture Methods

Section 8 Convolution and Deconvolution

Improvement Over General And Wider Class of Estimators Using Ranked Set Sampling

EGR 544 Communication Theory

MOSFET device models and conventions

Institute of Actuaries of India

The Central Limit Theorem

AdaBoost. AdaBoost: Introduction

EXISTENCE THEORY OF RANDOM DIFFERENTIAL EQUATIONS D. S. Palimkar

C(p, ) 13 N. Nuclear reactions generate energy create new isotopes and elements. Notation for stellar rates: p 12

Recursive Identification of Hammerstein Systems with Polynomial Function Approximation

PI3B V, 16-Bit to 32-Bit FET Mux/DeMux NanoSwitch. Features. Description. Pin Configuration. Block Diagram.

Sampling Example. ( ) δ ( f 1) (1/2)cos(12πt), T 0 = 1

King Fahd University of Petroleum & Minerals Computer Engineering g Dept

6.003 Homework #5 Solutions

International Association of Scientific Innovation and Research (IASIR) (An Association Unifying the Sciences, Engineering, and Applied Research)

Sampling. AD Conversion (Additional Material) Sampling: Band limited signal. Sampling. Sampling function (sampling comb) III(x) Shah.

Calculus BC 2015 Scoring Guidelines

MITPress NewMath.cls LAT E X Book Style Size: 7x9 September 27, :04am. Contents

Stochastic Processes Adopted From p Chapter 9 Probability, Random Variables and Stochastic Processes, 4th Edition A. Papoulis and S.

Calculus Limits. Limit of a function.. 1. One-Sided Limits...1. Infinite limits 2. Vertical Asymptotes...3. Calculating Limits Using the Limit Laws.

A quadratic convergence method for the management equilibrium model

Solutions Manual 4.1. nonlinear. 4.2 The Fourier Series is: and the fundamental frequency is ω 2π

Optimization-based Alignment for Strapdown Inertial Navigation System: Comparison and Extension

Learning in the Deep-Structured Conditional Random Fields

On Acoustic Radiation by Rotating Dipole Sources in Frequency Domain

6.003: Signals and Systems

Series Expansion with Wavelets. Advanced Signal Processing Reinisch Bernhard Teichtmeister Georg

Analysis of Using a Hybrid Neural Network Forecast Model to Study Annual Precipitation

Comparison between Fourier and Corrected Fourier Series Methods

Some Properties of Semi-E-Convex Function and Semi-E-Convex Programming*

Lecture 15 First Properties of the Brownian Motion

Mathematical Statistics. 1 Introduction to the materials to be covered in this course

Energy Density / Energy Flux / Total Energy in 1D. Key Mathematics: density, flux, and the continuity equation.

Extended Laguerre Polynomials

Moment Generating Function

Al Lehnen Madison Area Technical College 10/5/2014

Principles of Communications Lecture 1: Signals and Systems. Chih-Wei Liu 劉志尉 National Chiao Tung University

State-Space Model. In general, the dynamic equations of a lumped-parameter continuous system may be represented by

Optimum Receivers for the AWGN channel

Dynamic h-index: the Hirsch index in function of time

Numerical Method for Ordinary Differential Equation

O & M Cost O & M Cost

Chapter 3 Moments of a Distribution

Clock Skew and Signal Representation

International Journal of Multidisciplinary Approach and Studies. Channel Capacity Analysis For L-Mrc Receiver Over Η-µ Fading Channel

Adaptive Visual Servoing of Micro Aerial Vehicle with Switched System Model for Obstacle Avoidance

The analysis of the method on the one variable function s limit Ke Wu

A Generalized Cost Malmquist Index to the Productivities of Units with Negative Data in DEA

The universal vector. Open Access Journal of Mathematical and Theoretical Physics [ ] Introduction [ ] ( 1)

ELEG5693 Wireless Communications Propagation and Noise Part II

CHARACTERIZATIONS OF THE NON-UNIFORM IN TIME ISS PROPERTY AND APPLICATIONS

Supplement for SADAGRAD: Strongly Adaptive Stochastic Gradient Methods"

State and Parameter Estimation of The Lorenz System In Existence of Colored Noise

Hough search for continuous gravitational waves using LIGO S4 data

Procedia - Social and Behavioral Sciences 230 ( 2016 ) Joint Probability Distribution and the Minimum of a Set of Normalized Random Variables

Big O Notation for Time Complexity of Algorithms

12 Getting Started With Fourier Analysis

Math 6710, Fall 2016 Final Exam Solutions

Asymptotic statistics for multilayer perceptron with ReLu hidden units

BEST LINEAR FORECASTS VS. BEST POSSIBLE FORECASTS

Application of Intelligent Systems and Econometric Models for Exchange Rate Prediction

Stationarity and Error Correction

Effect of Heat Exchangers Connection on Effectiveness

ECE 340 Lecture 15 and 16: Diffusion of Carriers Class Outline:

CS623: Introduction to Computing with Neural Nets (lecture-10) Pushpak Bhattacharyya Computer Science and Engineering Department IIT Bombay

Lecture 8 April 18, 2018

EXAMPLE SHEET B (Partial Differential Equations) SPRING 2014

Fourier transform. Continuous-time Fourier transform (CTFT) ω ω

6/10/2014. Definition. Time series Data. Time series Graph. Components of time series. Time series Seasonal. Time series Trend

2 f(x) dx = 1, 0. 2f(x 1) dx d) 1 4t t6 t. t 2 dt i)

Hybrid System Modeling Using Impulsive Differential Equations

University of Southern California

Basic Results in Functional Analysis

Optimization of Rotating Machines Vibrations Limits by the Spring - Mass System Analysis

ES 330 Electronics II Homework 03 (Fall 2017 Due Wednesday, September 20, 2017)

6.003: Signal Processing Lecture 2a September 11, 2018

The Moment Approximation of the First Passage Time for the Birth Death Diffusion Process with Immigraton to a Moving Linear Barrier

So, given the value x generated by the computer, we solve this equation for y, which now has the

Inference of the Second Order Autoregressive. Model with Unit Roots

6.003: Signals and Systems Lecture 20 April 22, 2010

Math 2414 Homework Set 7 Solutions 10 Points

Statistical Estimation

6.003: Signal Processing Lecture 1 September 6, 2018

APPLICATION OF THEORETICAL NUMERICAL TRANSFORMATIONS TO DIGITAL SIGNAL PROCESSING ALGORITHMS. Antonio Andonov, Ilka Stefanova

Joint Spectral Distribution Modeling Using Restricted Boltzmann Machines for Voice Conversion

ECE 350 Matlab-Based Project #3

Solutions to selected problems from the midterm exam Math 222 Winter 2015

λiv Av = 0 or ( λi Av ) = 0. In order for a vector v to be an eigenvector, it must be in the kernel of λi

Transcription:

Paper Iroducio ~ Modellig he Uceraiy i Recoverig Ariculaio fro Acousics ~ Kori Richod, Sio Kig, ad Paul Taylor Tooi Toda Noveber 6, 2003

Proble Addressed i This Paper Modellig he acousic-o-ariculaory appig Speech producio Acousic speech sigal Ariculaory cofiguraio Iversio appig Acousic-ariculaory daa Saisical odel

Coes Iversio appig Acousic-ariculaory daa MOCHA Mappig wih ulilayer percepro MLP Mappig wih iure desiy ewor MDN Coparig MLP wih MDN

Iversio Mappig Mappig fro acousic speech sigal o ariculaory cofiguraio Ill-posed proble o-uique soluio Applicaios Speech codig Speech raiig Speech recogiio Speech syhesis

MOCHA daabase Mulichael ariculaory daabase MOCHA Quee Margare Uiversiy College Four daa sreas Acousic wavefor 16 Hz, 16 bi Larygograph 16 Hz, 16 bi Elecropalaograph Elecroageic ariculograph EMA 460 Briish TIMIT seeces 40 speaers Available: 2 speaers, ale ad feale

EMA: Elecroageic Ariculograph Saplig he ovee of receiver coils aached o he ariculaors 9 pois 2 referece pois Top lip, boo lip, boo icisor, ogue ip, ogue body, ogue dorsu, velu, bridge of he ose, upper icisor - ad y-coordiaes i he idsagial plae 14 chaels 500 Hz

Saples: 2-D plo 60 50 40 Bridge of he ose ref. 30 y [] 20 Upper icisor ref. Upper lip 10 Togue dorsu Togue body 0 Togue ip -10 Lower lip Velu -20 Lower icisor -30-20 -10 0 10 20 30 40 50 60 70 []

Saples: Tie sequece [] y [] 60 40 20 0-20 10 0-10 -20-30 0 200 400 600 800 1000 1200 Upper lip Velu Togue dorsu Togue body Togue ip Lower icisor Upper lip Togue ip Lower lip Togue dorsu Togue body Lower icisor Lower lip Velu 0 200 400 600 800 1000 1200 Tie [s]

MLP: Mulilayer Percepro Ipu layer: 400 uis 20 fraes of 20 filer ba coefficies Oupu layer: 14 uis 7 - ad y-coodiaes Hidde layer: 38 uis Pruig fro 50 uis Hidde layer Ariculaory oupu Full coecio Full coecio Acousic ipu

Feaures Acousic feaure 20 el-scale filerba coefficies 20 s haig widow, 10s shif Noralizaio: 1/4, 95% ierval [0.0, 1.0] Ariculaory feaure Lesseig he effec of oise caused by easuree error i EMA achie 10s shif Noralizaio: 1/5, 95% ierval [0.1, 0.9] Reovig silece fraes Daa se For raiig: 368 seeces For validaio: 46 seeces For es: 46 seeces

Weigh Opiizaio Error fucio = E y ; w N K y Gradie desce raiig w i+ 1 w i Scaled Cojugae Gradie SCG 2 ; w* = i = w + w = η E w i i w* a he iiu of E Usig o oly firs derivaives bu also secod derivaives

Eperieal Evaluaio Evaluaio easures Roo ea square RMS error Overall disace bewee wo rajecories Correlai coefficie Resuls Siilariy of shape ad sychroy of wo rajecories P. 7, Table 1 Average of RMS error: 1.62 Esiaed rajecories, p. 9, Figure 2

Shorcoigs of MLP Mappig Discoiuiy i esiaed rajecories Ariculaors ove slowly ad soohly Low-pass filerig Chael specific cuoff frequecy Isufficie o odel oe-o-ay appigs Usig coe widows Effecive bu isufficie Soe liiaios of usig he su-of-squares error fucio

MDN: Miure Desiy Newor Oupu: codiioal probabiliy desiy fucio Kerel: Gaussia fucio Weigh Mea Spherical covariace = M p φ α = M j j z z ep ep α α α µ µ z = ep σ σ z = 1 α 11 µ 1 σ 2 α 21 µ 2 σ p = K K K µ 2 /2 2 1 ep 2 1 σ σ π φ

Weigh Opiizaio i MDN Error fucio K-eas based iiializaio Ucodiioal desiy of he arge daa = M N E l φ α, z E π α α = =, 2 z E σ µ π µ = K z E 1, 2 σ µ π σ = = M j j j 1, φ α φ α π

Eperieal Evaluaio Resuls Oupu probabiliy, p. 14, Figure 4, 5 Low variace: accuracy is higher High variace: accuracy is lower Phoeic depede variaces, p. 16, Table 2. Low variace of criical ariculaor Togue ip y [s, z, ] Upper lip [, w, b, p] Ecepioal eaple Velu [,, ] Usig characerisics of variace

Coparig MDN wih MLP MLP Sigle Gaussia probabiliy desiy fucio MDN Mea: varied accordig o ipu oupu of MLP Variace: fied global variace i raiig daa Muliple Gaussia probabiliy desiy fucio Weigh: varied accordig o ipu Mea: varied accordig o ipu Variace: varied accordig o ipu

Copariso wih Mea Lielihood Resuls P. 18, Table. 3 Y-coordiaes is ore iproved ecep for velu Y: 13.3%, X: 4.8% Coparig probabiliy desiy fucios P. 19, Figure 6 Effeciveess of uli desiy for odellig oe-oay appigs

Coclusio Acousic-o-ariculaory appig Ill-posed proble Usig MOCHA daabase Speech wavefor EMA Mappig wih MLP or MDN MDN: ore fleible ad accurae odel Effecive o oe-o-ay appig Variace varied accordig o ipu sigal

Fuure Wor Coiuous rajecory Kala soohig wih variace Usig ariculaory iforaio Cos fucio i cocaeaive speech syhesis Specral esiaio Speech recogiio

Proposed Algorih Esiaed saic feaure Esiaed dyaic feaure 2 1 0-1 0.2 0.1 0-0.1-0.2 0 1000 2000 Tie [s] 1. Esiaig o oly saic bu also dyaic probabiliy Esiaed saic feaure 1.2 0.8 0.4 0-0.4 ML wih boh disribuios Saic ea 0 1000 2000 Tie [s] 2. Feaure geeraio wih ML for saic ad dyaic disribuios