Towards automated surfacelevel phonetic analysis of SL. Still cont d! Preamble



Similar documents
Evaluating the Finnish Sign Language Benchmark

Teaching Methodology for 3D Animation

ANALYZING A CONDUCTORS GESTURES WITH THE WIIMOTE

Sign language transcription conventions for the ECHO Project

Reading Specialist (151)

Overview The Corpus Nederlandse Gebarentaal (NGT; Sign Language of the Netherlands)

Basic Theory of Intermedia Composing with Sounds and Images

The Complexity of Lexical Movements in FinSL

HANDS-FREE PC CONTROL CONTROLLING OF MOUSE CURSOR USING EYE MOVEMENT

WinPitch LTL II, a Multimodal Pronunciation Software

Department of Music, University of Jyväskylä

A Method of Caption Detection in News Video

HSI BASED COLOUR IMAGE EQUALIZATION USING ITERATIVE n th ROOT AND n th POWER

Students with Reading Problems Their Characteristics and Needs

2 SYSTEM DESCRIPTION TECHNIQUES

Early Childhood Foundations to the Indiana Academic Standards

BLIND SOURCE SEPARATION OF SPEECH AND BACKGROUND MUSIC FOR IMPROVED SPEECH RECOGNITION

INCREASE YOUR PRODUCTIVITY WITH CELF 4 SOFTWARE! SAMPLE REPORTS. To order, call , or visit our Web site at

Introduction to Pattern Recognition

OCPS Curriculum, Instruction, Assessment Alignment

Comprehensive Reading Assessment Grades K-1

A Learning Based Method for Super-Resolution of Low Resolution Images

Program curriculum for graduate studies in Speech and Music Communication

American Sign Language and ASL Linguistics


Curriculum Vitae. Alison M. Trude December Website: Department of Psychology, University of Chicago

How Children Acquire Language: A New Answer by Dr. Laura Ann Petitto

Reading Competencies

Interactive Multimedia Courses-1

Sales Checkpoint Performance Feedback System

An Avatar Based Translation System from Arabic Speech to Arabic Sign Language for Deaf People

Learning Styles and Aptitudes

Text-To-Speech Technologies for Mobile Telephony Services

Monitoring Modality vs. Typology

LOCAL SURFACE PATCH BASED TIME ATTENDANCE SYSTEM USING FACE.

SignLEF: Sign Languages within the European Framework of Reference for Languages

READING SPECIALIST STANDARDS

American Standard Sign Language Representation Using Speech Recognition

A Cognitive Approach to Vision for a Mobile Robot

The syllable as emerging unit of information, processing, production

Revisiting the readability of management information systems journals again

False alarm in outdoor environments

Visibility optimization for data visualization: A Survey of Issues and Techniques

High-Mix Low-Volume Flow Shop Manufacturing System Scheduling

Online Catalogue

GLOVE-BASED GESTURE RECOGNITION SYSTEM

Manual signs: what does research tell?

Effects of Gender and Word Choice on Qualitative Inference During Reading

LECTURE 11: PROCESS MODELING

PRE AND POST TEST TO SEE THE DIFFERENCE BETWEEN YEARS OF ANIMATED LITERACY AND KNOWLEDGE OF LETTERS STEPHANIE, BUCK. Submitted to

SLDTC: The Sign Language Documentation Training Center

PHYSIOLOGICALLY-BASED DETECTION OF COMPUTER GENERATED FACES IN VIDEO

NAVIGATING SCIENTIFIC LITERATURE A HOLISTIC PERSPECTIVE. Venu Govindaraju

Open Access A Facial Expression Recognition Algorithm Based on Local Binary Pattern and Empirical Mode Decomposition

2) Language: Lesion Studies Demonstrating the Left Hemisphere s Dominance for Language - Broca s Aphasia, Wernicke s Aphasia

KNOWLEDGE ORGANIZATION

Winter 2016 Course Timetable. Legend: TIME: M = Monday T = Tuesday W = Wednesday R = Thursday F = Friday BREATH: M = Methodology: RA = Research Area

Linear Programming for Optimization. Mark A. Schulze, Ph.D. Perceptive Scientific Instruments, Inc.

An Introduction to Online Teaching and Learning. Good teaching is good teaching, no matter how it s done.

SWING: A tool for modelling intonational varieties of Swedish Beskow, Jonas; Bruce, Gösta; Enflo, Laura; Granström, Björn; Schötz, Susanne

A Database Tool for Research. on Visual-Gestural Language

Final Year Projects at itm. Topics 2010/2011

CURRICULUM VITAE. Jessica Sari Fleming Hay. Department of Psychology University of Tennessee, Knoxville 1404 Circle Drive Knoxville, TN 37996

Detection and Restoration of Vertical Non-linear Scratches in Digitized Film Sequences

The plot is the sequence of events in a story. Each event causes or leads to the next. Events of the plot reveal a problem called the conflict.

Particles, Flocks, Herds, Schools

Face Identification by Human and by Computer: Two Sides of the Same Coin, or Not? Tsuhan Chen

Semantic Video Annotation by Mining Association Patterns from Visual and Speech Features

USABILITY OF A FILIPINO LANGUAGE TOOLS WEBSITE

Visual Storytelling, Shot Styles and Composition

KIN Biomechanics LAB: Ground Reaction Forces - Linear Kinetics

Locating and Decoding EAN-13 Barcodes from Images Captured by Digital Cameras

Sign Language Linguistics Course texts Overview Assessment Week 1: Introduction and history of sign language research

Chapter 7: Structuring System Process Requirements

Open Source UAS Software Toolkits. Keith Fieldhouse Technical Lead, Kitware Inc.

Quality Estimation for Scalable Video Codec. Presented by Ann Ukhanova (DTU Fotonik, Denmark) Kashaf Mazhar (KTH, Sweden)

Template-based Eye and Mouth Detection for 3D Video Conferencing

Making Machines Understand Facial Motion & Expressions Like Humans Do

An Automated Analysis and Indexing Framework for Lecture Video Portal

Object tracking & Motion detection in video sequences

Research on Animation Lip synchronization technology

Laser Gesture Recognition for Human Machine Interaction

How To Filter Spam Image From A Picture By Color Or Color

Potential of face area data for predicting sharpness of natural images

Name Partners Date. Energy Diagrams I

LABORATORY TWO GEOLOGIC STRUCTURES

W-PhAMT: A web tool for phonetic multilevel timeline visualization

Tutorial for Tracker and Supporting Software By David Chandler

Announcements. SE 1: Software Requirements Specification and Analysis. Review: Use Case Descriptions

Software Requirements Specification

Sing, Spell, Read & Write

Sign Language Phoneme Transcription with Rule-based Hand Trajectory Segmentation

ISSN: A Review: Image Retrieval Using Web Multimedia Mining

jorge s. marques image processing

Translog-II: A Program for Recording User Activity Data for Empirical Translation Process Research

Investigating tongue movement during speech with ultrasound

COMPUTING CLOUD MOTION USING A CORRELATION RELAXATION ALGORITHM Improving Estimation by Exploiting Problem Knowledge Q. X. WU

Video in Logger Pro. There are many ways to create and use video clips and still images in Logger Pro.

Science, Technology, Engineering & Mathematics Career Cluster

An Overview of the Developmental Stages in Children's Drawings

Transcription:

Presentation at the Brown Bag seminar (JyU) on 20 April 2009 Pre-preamble Towards automated surfacelevel phonetic analysis of SL From 2007 onwards: An attempt to establish a multidisciplinary research project in which computer vision techniques for the recognition and analysis of gestures and facial expressions from video are developed and applied to the processing of Sign Language in general and Finnish Sign Language in particular. Tommi Jantunen, Department of Languages, University of Jyväskylä tommi.j.jantunen@jyu.fi Cont d Still cont d! Project partners: Helsinki University of Technology: Jorma Laaksonen & Markus Koskela (CIS) University of Jyväskylä: Ritva Takkinen & Tommi Jantunen (Dept. of Languages); Timo Ahonen & Auli Meronen (NMI) University of Art and Design: Antti Raike (Media Lab) Finnish Association of the Deaf: Päivi Rainò (Sign Language Unit) Publication: Koskela, Markus; Laaksonen, Jorma; Jantunen, Tommi; Takkinen, Ritva; Rainò, Päivi & Raike, Antti (2008). Contentbased video analysis and access for Finnish Sign Language - a multidisciplinary research project. In O. Crasborn, E. Efthimiou, T. Hanke, E. D. Thoutenhoofd & I. Zwitserlood (Eds.), Construction and exploitation of sign language corpora [Proceedings of the 3rd Workshop on the representation and processing of sign languages, organised as a part of the 6th Language resources and evaluation conference (LREC) at Marrakech, Morocco, June 1st, 2008], pp. 101-104. Paris: ELRA. Preamble Cont d From 2009 onwards: A smaller scale multidisciplinary project aiming at representing graphically, from the existing video material, the prosody and rhythm of continuous natural signing My own work on signed syllable, sign, and sentence (cf. the need for more empirical research) Päivi s project (OSATA) dealing with dyslectic signers (cf. the role of rhythm in explaining dyslexia) CIS s interest in testing content-based video retrieval and analysis methods (cf. CIS s celebrated PicSOM technology) A research group consisting of experts in Computer and information science: Jorma Laaksonen & Markus Koskela (HUT/CIS) Sign language linguistics: Tommi Jantunen (JyU) & Päivi Rainò (FAD) Spoken language phonetics and prosody: Eeva Yli-Luukko ( Kotus ), Eija Aho (University of Helsinki) & Richard Ogden (University of York) 1

Introduction (1) Background: In spoken language phonetics, representing speech data directly with different graphical diagrams is quite widespread, or at least not uncommon (cf. the use of Praat etc.). Graphical representation of data enables more accurate analysis. Introduction (2) Current problem : There exists a number of SL studies that claim to be phonetic in a similar way spoken language phonetic studies are. However, in the literature, there are only a handfull of papers in which data is represented directly in a form of a graphical diagram (cf. the lack of movement tracking hardware/software etc.) When compared to the phonetic analysis of spoken language, the analysis of SL is based relatively more on estimates and abstractions. Motion tracking systems (1) My Wiimote experiment Motion tracking systems have been used in the study of SL phonetics (cf. movement) at least since Wilcox (1992) These systems are nowadays VERY accurate and enable graphical representation of the data as well as the extraction of various types of information from the data Cont d Still cont d! 2

Motion tracking systems (2) However, motion tracking data is ALWAYS laboratory-data Method(olog)ical prerequisite for natural data (in the sense of prosody and function ) is pre-recording! Is it possible to graphically represent and analyze SL (cf. movement) purely on the basis of video s digital content? One rare example from the existing literature Boyes Braem, Penny (1999). Rhythmic temporal patterns in the signing of deaf early and late learners of Swiss German Sign Language. Language and Speech 42:2-3, 177-208. Boyes Braem (1999:189) Boyes Braem (1999:188) DEMO 1 How to graphically represent and analyse SL movement on the basis of digital content of existing videos? Suvi 1038/3 3

HUT s 2007 demo An example frame Koskela & al. (2008:103) [A]n essential feature in the analysis of recorded continuous-signing sign language is that of motion. For tracking local motion in the video stream, we apply a standard algorithm based on detecting distinctive pixel neighborhoods and then minimizing the sum of squared intensity differences in small image windows between two successive video frames [ ]. DEMO 2 Towards more sophisticated content-based analysis of SL videos Skin filter Improved vector count 4

Auto-calculated values The result number of tracked motion points horizontal motion vertical motion length of sum of motion vectors sum of motion vector lengths length of sum of acceleration sum of acceleration lengths Horizontal motion Observations (Suvi 1038/3, hm) In general, changes in horizontal motion map well to the phonological boundaries (cf. the method of identifying lexical signs) However, the boundaries are not unambiguous in all cases Horizontal motion of lexical sequences exhibits more variation than that of intersign transitions (cf. the trad. assumption that only lexical movements are modifiable) The amount of horizontal motion in reduplicated or iterated signs reduces towards the end (cf. the disyllabicity constraint ) Vertical motion Observations (Suvi 1038/3, vm) Cf. the previous observations concerning horizontal motion Note that the vertical motion during lexical signs is directed from top to bottom, or is a plateau Note that the first two signs (BOY and INDEX) are produced with a continuous downward movement 5

Number of tracked motion points Observations (Suvi 1038/3, ntmp) Radical changes in the number of motion tracked interest points occur at the sign/transition boundaries Sequences with maximal amount of motion points seem to be transitions, or short sequences centering around sign/transition boundaries Note the large number of mtips at the end of the compound COMPUTER; also the small amount of mtips during the last two signs Note the main levels of mtips (cf. the number of signs) Length of sum of motion vectors Sum of motion vector lengths Observations (Suvi 1038/3, lsmv&smvl) Peaks map to transitions and to sign/ transition boundaries, not to lexical signs per se The lowest values occur in signs HOBBY and PLAY-JOYSTIC (cf. the muscular tension in the production of these signs) In general, movement in all tracked points occurs into same direction Length of sum of acceleration 6

Sum of acceleration lengths Observations (Suvi 1038/3, lsa&sal) Acceleration peaks map to sign/ transition boundaries and to transitions Acceleration values within lexical signs are relatively lower than the values within transitions Pessi & Illusia (frames1-512) DEMO 3 An experiment with a longer story Number of tracked motion points (P&I, frames 1-10705) Aho & Yli-Luukko (2005:209; laajoja intonaatiojaksoja eteläpohjalaisen naisen kertomuksessa) 7

DEMO 4 A few words concerning the qualitative difference between sign internal content movements and transitional movements Acceleration curve of one sentence in the P&I story Acceleration curve of a lexical sign in P&I (1) Acceleration curve of a lexical sign in P&I (2) Acceleration curve of a lexical sign in P&I (3) Acceleration curve of a transition in P&I (1) 8

Acceleration curve of a transition in P&I (2) Acceleration curve of a transition in P&I (3) Acceleration > perception > sonority [A] visual beat is communicated by periods of acceleration or deceleration [ ] (Luck & Sloboda 2008:237) Visual beats = those events that are felt to be more forcefully produced and around which the other events in the sequence are organized (Allen & al. 1991:197) Sonority is perceptual salience (e.g. Ohala 1990) Consequences to phonological theory? The axiom of signed syllable research has been that, in a sign stream, the most salient/sonorous events associate to lexical signs However, the present data suggests that this is not the case! Issues that still need addressing The Z-dimension (cf., for example, the sign BOY in Suvi s example 1038/3) Symmetrical two-handed signs (cf. the Pessi & Illusia story) Option to choose the focus area References Aho, Eija & Yli-Luukko, Eeva (2005). Intonaatiojaksoista. Virittäjä 2/2005, 201-220. Allen, George D.; Wilbur, Ronnie B. & Schick, Brenda B. (1991). Aspects of rhythm in ASL. Sign Language Studies 72, 297-320. Boyes Braem, Penny (1999). Rhythmic temporal patterns in the signing of deaf early and late learners of Swiss German Sign Language. Language and Speech 42:2-3, 177-208. Koskela, Markus; Laaksonen, Jorma; Jantunen, Tommi; Takkinen, Ritva; Rainò, Päivi & Raike, Antti (2008). Content-based video analysis and access for Finnish Sign Language - a multidisciplinary research project. In O. Crasborn, E. Efthimiou, T. Hanke, E. D. Thoutenhoofd & I. Zwitserlood (Eds.), Construction and exploitation of sign language corpora [Proceedings of the 3rd Workshop on the representation and processing of sign languages, organised as a part of the 6th Language resources and evaluation conference (LREC) at Marrakech, Morocco, June 1st, 2008], pp. 101-104. Paris: ELRA. Luck, Geoff & Nte, Sol (2008). An investigation of conductors' temporal gestures and conductormusician synchronization, and a first experiment. Psychology of Music 36:1, 81-99. hala, J. (1990). Alternatives to the sonority hierarchy for explaining segmental sequential constraints. In M. Ziolkowski, M. Noske & K. Deaton (Eds.), CLS 26. [Papers from the 26th regional meeting of the Chicago Linguistic Society] Vol. 2, The parasession on the syllable in phonetics and phonology, 319-338. Chicago Linguistic Society, University of Chicago, Chicago, Ill. Wilcox, Sherman (1992). The phonetics of fingerspelling. Studies in speech pathology and clinical linguistics 4. Amsterdam: John Benjamins. 9