An animal’s flow of behavior often appears to be segmented into elementary units. Parsing the behavior by setting some kind of a cut-off point might yield units that agree with the observer’s intuition, but do not guarantee a classification according to natural morphogenetic units. We look for a division that is dictated by the geometrical and statistical properties of the data, rather than by some ad hoc choice. One strategy that we use is to define a numerical measure of motion, in which we check for the existence of a natural threshold value T. By this we mean that the set of values bigger than T has a statistical distribution markedly different from that of the values smaller than T. This strategy will be illustrated by the way we parse a rat’s path into "stops" and "progressions".

At the level of body segments, the problem is to define and measure "action patterns" in terms of the complex coordination between the movements of all the segments, as well as parse them in time. Our strategy is to treat the movement at every point in time as part of a local cyclic movement, and measure the phase, amplitude and frequency of this cycle [1]. The relative phase (i.e., the phase difference) between concurrent movements is a good candidate for defining and recognizing the action pattern, while discontinuities in the amplitude and frequency mark the boundary in time between specific movements. For this purpose we developed special algorithms, and they will be demonstrated on ferret free locomotion.

This research was supported by the Israel Science Foundation funded by the Israel Academy of Sciences and Humanities.

References

1. Kafkafi, N.; Golani, I. (1998). A traveling wave of lateral movement coordinates both turning and forward walking in the ferret. Biological Cybernetics, in press.