TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
Journée Inter-GDR ISIS et Robotique: Interaction Homme-Robot
1. Model of expressive gestures
for humanoid robot NAO
Quoc Anh Le
Telecom ParisTech
Catherine Pelachaud
CNRS, Telecom ParisTech
2. NAO robot
• An autonomous, programable and medium-size humanoid robot (57 cm)
• Made by a French company (Aldebaran Robotics, Paris)
• 25 degrees of freedom
• 2 speakers to speak or play sound files
• Control colors of its eyes
3. Introduction
• This work is part of the National project ANR
GVLEX
– Partners: Aldebaran (coordinator), Acapela, LIMSI
and Telecom-ParisTech.
• Objective: Have the humanoid robot, Nao, read
story agreeably through expressive verbal and
non-verbal behaviors.
• Methodology: Select and compute behaviors
based on information extracted from the story:
its structure, its various semantic and pragmatic
elements as well as its emotional content.
• Ongoing work at the Telecom-ParisTech:
Control gestures of Nao by using the platform of
an existing virtual agent system, Greta
3
4. Greta platform
• Following SAIBA framework
• Two representation languages:
– FML: Function Markup Language
– BML: Behavior Markup Language
4
5. SAIBA framework
• Goal: Define interfaces at separate levels of abstraction for
multimodal behavior generation
• Structure: consist of 3 separated modules
– Intent planning: Planning of a communicative intent
– Behavior planning: Planning of multimodal behavior that carry
– Behavior realization: Realization of the planned behaviors
5
6. FML
• Objective: Encode communicative and expressive intent what agent aims to
transmit. It may be emotional states, beliefs or goals, etc.
6
7. BML
• Objective: Specify multimodal behaviors with constrains to be realized by
agent
• Description:
1. Occurrence of behaviors
2. Relative timings of behaviors
3. Form of behaviors/Reference to predefined animations
4. Conditions/Events/Feedbacks
2
1
3
7
8. Steps
1. Build a repertoire of gestures based on a video corpus
2. Use Greta platform to compute gestures for robot
Behavior
Realizer
BML
FML
Text Intent Behavior Planning
Planning
Behavior
BML Realizer
9. Build repertoire of gestures
• Goal: Collect expressive gestures of individuals
in a specified context (story-tellers)
• Stages:
1. Video collection
2. Code schema and annotations
3. Elaboration of symbolic gestures
annotations Gesture elaboration
Videos Gesture
Editor
corpus Repertoire
9
10. Video collection
• 6 actors from an amateur troupe were
videotaped
• Actors had received the script of the story
beforehand
• The text was displayed during the session so
that they could read it from time to time
• 2 digital cameras were used (front and side-
view)
• Each actor was videotaped twice
– 1st session as a training / warm-up session
– the most expressive session can be kept for analysis
10
Martin
12. Code schema and annotation
• Code schema
– Goal: enable specification of gesture lexicons for Greta and Nao
– Segmentation based on gesture phrases
– Attributes
• Handedness : Right hand / Left hand / 2 hands
• Category: deictic, iconic, metaphoric, beat, emblem
(McNeill 05, Kendon 04)
• Lexicon: 47 different entries
• Annotations using Anvil tool (Kipp 01)
– Current state: 125 gestures segmented for 1 actor
– Rich in terms of gestures : 23 gestures per minutes for subject
12
Martin
16. Predefined positions
• Pre-calculate joint values of
all combinations of hand
positions in 3D space
(vertical, horizontal, distance)
using Choregraphe.
• Current state: 105 positions
corresponding to 7 vertical
values, 5 horizontal values
and 3 distance values (000,
001, … , 462).
• Replace symbolic positions
by real joint values when
compiling. 16
18. Behavioral scripts using
ChoreGraphe
Voilà bien longtemps, un soir de printemps, trois petits morceaux de nuit se
détachèrent du ciel et tombèrent sur Terre….
18
19. Gesture lexicon
• Different degrees of freedom
• Variant of a gesture encompasses a family of gestures that shares
– the same meaning (e.g. to stop someone)
– a core signal (e.g. vertical flat hand toward the other)
• Gestures within a family may differ along the non-core signals they
use
• Construction of a common lexicon with
– Greta-Gestuary
– Nao-Gestuary
• In the specific lexicon, variant shares similar meaning and signal-
core.
19
20. Robot vs. Greta
• Degree of freedoms
• Not dynamic wrists
• Three fingers that open or close together
• Movement speed (>0.5 seconds)
• Singular positions
=> Gestures may not be identical but should convey similar meaning
20
25. Future work
• Synchronization of gestures with speech
for robot
• Define invariant signification of gestures
• Define different levels of BML descriptions
for gestures
• Define evaluation methods
25