Top Banner
Handrix: Animating the Human Hand George ElKoura Karan Singh George ElKoura Karan Singh University of Toronto University of Toronto
39

Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Dec 16, 2015

Download

Documents

Bénédict Paul
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Handrix: Animating the Human Hand

George ElKoura Karan SinghGeorge ElKoura Karan Singh

University of TorontoUniversity of Toronto

Page 2: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.
Page 3: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Motivation

How do we get from a piece of music:How do we get from a piece of music:

To finger motions that play the music:To finger motions that play the music:

Page 4: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Overview

• System Architecture OverviewSystem Architecture Overview

• Guitar AlgorithmGuitar Algorithm

• Anatomy of the HandAnatomy of the Hand

• Hand ModelHand Model

• ResultsResults

Page 5: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

System Architecture

Page 6: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

System Architecture

Page 7: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

System Architecture

Page 8: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

System Architecture

Page 9: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

System Architecture

Page 10: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Related work

• Anatomy Anatomy

– HHäägerger-R-Ross and oss and SSchieberchieber,, 200 20000

• VisionVision

– Lee and Kunii, 1995Lee and Kunii, 1995

• Graphics Graphics

– Rijpkema and GRijpkema and Girardirard, 1991, 1991

• MusicMusic

– Quine, 1990Quine, 1990

Page 11: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

A Bit of Anatomy

• Why are fingers interdependent?Why are fingers interdependent?

– ““One to many” muscle insertion sitesOne to many” muscle insertion sites

– Close proximity of tendonsClose proximity of tendons

– Neurological constraintsNeurological constraints

• A clear anatomical understanding is still A clear anatomical understanding is still being developedbeing developed

Page 12: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Hand Model

• Goal:Goal:

– Map partially specified hand positions to realistic hand posturesMap partially specified hand positions to realistic hand postures

• Approach:Approach:

– Motion Capture lots of hand positionsMotion Capture lots of hand positions

– Nearest Neighbors, RBFs to approximate hand-pose spaceNearest Neighbors, RBFs to approximate hand-pose space

                                        

                                      

Page 13: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Assumptions• Assumptions to reduce number of parameters:Assumptions to reduce number of parameters:

1.1. Independence of the ThumbIndependence of the Thumb

2.2. Invariance of Hand DominanceInvariance of Hand Dominance

3.3. Independence of Adduction/AbductionIndependence of Adduction/Abduction

4.4. Irrelevance of FrequencyIrrelevance of Frequency

5.5. Locality: Posture of wrist and arm do not affect digit Locality: Posture of wrist and arm do not affect digit independence independence

• These assumptions reduce the problem to 12 DOFs.These assumptions reduce the problem to 12 DOFs.

Reference: Häger-Ross and Schieber, Journal of Neuroscience 20(22), 2000

© Babette Adrian

Page 14: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Reconstruction

• Allow the animator/controller to partially specify Allow the animator/controller to partially specify a hand posture (a hand posture (00, , 11, … , … nn))

Page 15: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Reconstruction

• Allow the animator/controller to supply weights Allow the animator/controller to supply weights for each joint indicating the importance of the for each joint indicating the importance of the supplied joint angle (supplied joint angle (00, , 11, …, , …, nn))

0 = 11

n = 00

Page 16: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Reconstruction

• Correct the user-supplied hand posture using the Correct the user-supplied hand posture using the given weights and the database of physically given weights and the database of physically possible posturespossible postures

Page 17: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Reconstruction Method• The database is a collection of joint angles represented by 12D vectors: The database is a collection of joint angles represented by 12D vectors: II

• Compute the weighted distance by:Compute the weighted distance by:

• The The kk-Nearest Neighbors interpolation weights are computed by:-Nearest Neighbors interpolation weights are computed by:

Page 18: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Results(No Reconstruction)

Page 19: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Results(With Reconstruction)

Page 20: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Interdependence Results(Comparison)

Page 21: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

The Guitar

6 strings numbered from 1 to 6 6 strings numbered from 1 to 6 (heavy to light)(heavy to light)

Frets delimit strings lengths Frets delimit strings lengths producing audibly different notesproducing audibly different notes

Frets are numbered in increasing Frets are numbered in increasing order from the nut (fret 0)order from the nut (fret 0)

Page 22: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Tablature

• Common guitar music notationCommon guitar music notation

• Tells musician what frets to holdTells musician what frets to hold

• Does not tell musician what fingers to useDoes not tell musician what fingers to use

• Simple ASCII, readily available (OLGA) Simple ASCII, readily available (OLGA)

• Adaptable to specifying multi reaching goals over Adaptable to specifying multi reaching goals over time time

Page 23: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Fretting Design Objectives

1. Economy of effort1. Economy of effort

2. Read-ahead control2. Read-ahead control

3. Maintain natural motion of fingers3. Maintain natural motion of fingers

4. Capture differences in finger dexterity and 4. Capture differences in finger dexterity and strengthsstrengths

5. Fingertips touch guitar at the correct fret5. Fingertips touch guitar at the correct fret

Reference: Hector Quine, Guitar Technique, Oxford University Press, 1990

Page 24: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Guitar Space

Page 25: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Cost Functions

• Assign a different cost function to each finger, wristAssign a different cost function to each finger, wrist

• Each cost function can be a sum of specific cost functions to Each cost function can be a sum of specific cost functions to separately handle dexterity, strength and so forthseparately handle dexterity, strength and so forth

• Overall cost is computed, and cheapest fingering winsOverall cost is computed, and cheapest fingering wins

Page 26: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Finger Resolution Algorithm

Page 27: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Finger Collisions

Finger collisions are handled in a simple wayFinger collisions are handled in a simple way

• When a collision is detected, the colliding finger is moved When a collision is detected, the colliding finger is moved backwards on the fretbackwards on the fret

Page 28: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – C-Major Scale

Page 29: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – Chromatic Scale (fails)

Page 30: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – Chromatic Scale (passes)

Page 31: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – Chord Transitions

Page 32: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – More Chords

Page 33: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – Bar Chords

Page 34: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – C-Major Scale (3 Fingers)

Page 35: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Results – Stairway to Heaven

Page 36: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Future Work

Apply to other areas:Apply to other areas:

Page 37: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Future Work

• Slides, bends, hammersSlides, bends, hammers

• Frequency of motionFrequency of motion

• Motion Dynamics (strumming hand)Motion Dynamics (strumming hand)

• Anatomical Hand ModelAnatomical Hand Model

• Extension to other areas (e.g. Extension to other areas (e.g. locomotion)locomotion)

Page 38: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Summary

System Architecture OverviewSystem Architecture Overview

Guitar AlgorithmGuitar Algorithm

Anatomy of the HandAnatomy of the Hand

Hand ModelHand Model

ResultsResults

For More Information Please Visit:For More Information Please Visit:

http://www.dgp.toronto.edu/~gelkoura/handrix/index.html

Page 39: Handrix: Animating the Human Hand George ElKoura Karan Singh University of Toronto.

Acknowledgements

DGP Lab, University of TorontoDGP Lab, University of Toronto

Chris LandrethChris Landreth

Dave BaasDave Baas

Side Effects Software, Inc.Side Effects Software, Inc.

Alias SystemsAlias Systems

NSERCNSERC

For more information please visit:

http://www.dgp.toronto.edu/~gelkoura/handrix/index.html