Biological movement is built up of sub-blocks or motion primitives. Such primitives provide a compact representation of movement which is also desirable in robotic control applications. We analyse handwriting data to gain a better understanding of use of primitives and their timings in biological movements. Inference of the shape and the timing of primitives can be done using a factorial HMM based model, allowing the handwriting to be represented in primitive timing space. This representation provides a distribution of spikes corresponding to the primitive activations, which can also be modelled using HMM architectures. We show how the coupling of the low level primitive model, and the higher level timing model during inference can produce good reconstructions of handwriting, with shared primitives for all characters modelled. This coupled model also captures the variance profile of the dataset which is accounted for by spike timing jitter. The timing code provides a compact representation of the movement while generating a movement without an explicit timing model produces a scribbling style of output.
The theory of Optimal Unsupervised Motor Learning shows how a network can discover a reduced-order controller for an unknown nonlinear system by representing only the most significant modes. Here, I extend the theory to apply to command sequences, so that the most significant components discovered by the network correspond tomotion "primitives". Combinations of these primitives can be used to produce a wide variety of different movements. I demonstrate applications to human handwriting decomposition and synthesis, as well as to the analysis of electrophysiological experiments on movements resulting from stimulation of the frog spinal cord. 1 INTRODUCTION There is much debate within the neuroscience community concerning the internal representationof movement, and current neurophysiological investigations are aimed at uncovering these representations. In this paper, I propose a different approach that attempts to define the optimal internal representation in terms of "movement primitives", and I compare this representation with the observed behavior.
These primitives could be similar to primitives of computer science in that complex programs of behavior repertoires could be built from them. The biological primitives could thus form a bootstrap or basis set for learned motor behaviors. The concept of primitives used here is closely allied to other ideas in neurophysiology, notably the compounding of reflexes suggested by Sherrington, the building blocks of Getting, and the flexor reflex afferent ideas of Lundberg, Jankowska and colleagues. It has long been known that the spinal frog (i.e. This panoply of behavior demonstrates a local capacity of the isolated spinal cord for organizing adjusted action sequences.
Next we describe our use of human movement data as input and the huinanoid simulation test-bed for evaluation. We follow with a detailed discussion of three means of deriving primitives, the key component of our model, and describe implementations for each of them, as well as experimental results, demonstrated using human movement, captured with vision or magnetic markers, and imitated on a humanoid torso with dynamics, performing various movements from dance and athletics. Motivation Humanoid robots will increasingly become a part of human everyday lives, as they are introduced as caretakers for the elderly and disabled, assistants in surgery and rehabilitation, and educational toys for children. To make this possible, the process of robot programming and control nmst be simplified, and human-robot interaction must be made more natural. Our approach to addressing these challenges is to use biologically inspired notions of behavior-based control, and endow robots with the ability to imitate, so that they can be programmed and interacted with through human demoustration, a natural huxnan-humanoid interface.
We discuss the motion-based gestural primitives. Unlike the problems in facial expressions or communicative gestures which are subjected to the agreeable interpretation of "what it means", our problems are raised in the development of a human-machine performance system. We put an emphasis on dynamics and kinetic elements among interacting entities that comprise a complex environment for Human-Machine performance. In that environment the performers actively change the sensory information by means of movements; the performance movements that are the changes of the states of the performers induce the changes in system states. The paper introduces the concepts of synthesis of interactivity, extended circularity, and tonmeister kinesthetic. Gestural primitives are classified into three classes with respect to performer's orientation; trajectory-based, force-based, and pattern-base primitives. Introduction: Human-Machine Performance The testbed for this study is based on an architecture for human-machine performance. Human-machine performance is an active observation task initiated by a human observer in an environment where 1) divisions of labor between human and machine are well defined, 2) the observation task is time-critical, supported by synchronous feedback, and 3) interaction is assisted with various means to enhance comprehension of the behavior of mechanisms under exploration. The implementation of the architecture is guided by criteria for enabling timecritical observation. We provide the capability for every process to be defined having an independent period of execution. Quality of service specifications define synchronization as a distributed measure applied between each pair of parallel processes that directly service one another.