Database Reference
In-Depth Information
Economic and feasible storage of motion data by a computer,
•
Recognition of a sequence of motion coordinates,
•
Defi nition, orthography and morphology of motion dialects.
•
Movement notation and sign languages for the deaf constitute fi rst attempts at “motion
reading.” The fl exible notation system called the Laban Notation (LN), invented in 1928 by
Rudolf Laban, was at fi rst applicable only to musical partita, but later, it was improved into
what is today labeled the “Cinematography Laban”. In 1958 Noah Eshkol and Abraham
Wachman published a notation for describing movement based on a geometric concept
(Eshkol & Wachmann, 1958; Hutchinson, 1960).
Further developments came about during efforts to make use of notations for the purpose
of representing motion on the computer. Badler and Smoliar (1979) adopted parameters from
Laban's notation in developing a motion simulation system using the “key frames” approach.
However, neither the key frames approach nor indeed any model based on key positions,
including the LN notation, fully refl ects the phenomenon of motion. These approaches are
all based on the following assumptions:
1.
Motion is a derivative of
positions
. The transition between two sequential positions
proceeds at a uniform rate, and through the shortest possible route.
2.
Every change of position in space made by a limb refl ects movement performed by that
same limb. In other words, no distinction is made between movements and dragging
and between movements and compensations. Dragging is a passive motion that occurs
when a given limb changes its position without changing the internal geometric rela-
tion with adjacent links, e.g., the arm moves as a result of motion by the forearm.
We question these assumptions and suggest that they may be relaxed in defi ning the
basic building blocks of motion.
Among the limitations of the present motion notations are a lack of tools for storing and
processing the motion “text” and limited accuracy in expressing time and space (Bruderlin &
Williams, 1995; Calvert & Chapman, 1982; Earnshaw, Mangnenat-Thalmann, Terzopoulos
& Thalmann, 1998; Hodgins, Wooten, Brogan & O'Brien, 1995; Ko & Badler, 1996; van de
Pannw, 1996). These limitations stem from the fact these notations are basically
documenta-
tion
rather than storage and analysis tools.
These limitations make it diffi cult to effi ciently store and process motion signs and
symbols, and to analyze them by means of computer systems. To do this we need to defi ne
basic motion building blocks that can pave the way to a comprehensive motion language that
has an internal consistency as is common in any language. Recent work focuses on linguistic
aspects of the control of robots by means of a motion description language (Egerstedt, 2001).
A survey of issues and challenges in motion modeling is given in Agarwal et al. (2002).
In addition to the problem of representation, there is the issue of storage. Motion
data is usually attained through a “tracking system” that chooses single motion units at a
frequency of up to 5 Khz. Such a database very quickly becomes intolerably large. Thus,
for example, the data required to store the movement of 22 links (as with the Life Forms
system; see below) takes approximately
1.21 MBits per second
, and this is for representing
a
fi gures frame only
:
