Goals for transforming and decompressing a representation of & for action

The primary goal is success in the framework of evolution. More specific goals should help survival and propagation.

Let us start with the concept of a coherent action sequence

• A kitten or a young child might be chasing a ball, see a butterfly, and start chasing the butterfly
• Chasing a ball, or a butterfly, is a coherent action sequence. Changing from chasing the ball to chasing the butterfly shows the boundary, and the transition from one sequence to another
• Coherent or 'purposive' action sequences can vary in length and in changes required (non-repeatability of one segment to the next)
• Chasing a prey is a good example, where many unpredictable turns may be required
• Coherent or 'purposive' action sequences may not be precisely repeatable
• Chasing a prey on one occasion we may have to turn left then right. On the next occasion it may require a turn around a tree or some other obstacle that makes the path quite different from the preceding chase.

Next, let us look at action continuity and coordination

• To walk, we move forward on one leg, so that the leg is initially in front of us and then behind us. At the same time the other leg is swinging slightly above the ground from behind us to in front of us. At the end of these two coordinated swings we shift weight from one leg to the other.
• At the same time, synchronously, the arms may swing at our sides
• At the same time, synchronously, the hips may swivel slightly, typically up on the supporting side and down on the freely swinging side
• At the same time, synchronously, the hips may rotate slightly
• At the same time, synchronously, the head may move up and down, the hair may swing side to side, etc.
• Perception of the stride may allow us to discriminate between males and females, young and old, etc.
• Individual characteristics of the stride may help us to recognize particular individuals

Next, let us look at action specification to the muscles

• We talked about 1 msec interval between sets of instructions to the muscles
• It would be easier if the instructions did not change during the stride
• Because of changes in body balance, it seems very unlikely that the instructions are invariant even during a single stride

Next, let us look at action perception through vision

• We talked about 18 frames per second, i.e. 55 msec intervals between successive images
• Because of changes in relative angles between legs, between arms, and in hip movement, each frame is likely to diifer from the preceding frame

Next, let us look at a single stride in a fairly fast walk, i.e. where the weight is on the right foot, and where the stride lasts for half a second

• This corresponds to 500 instructions and at least 9 visual images, i.e. about 55 instructions per image.
• Even actors, having seen moving pictures of the person they are imitating, probably cannot memorize that level of detail about the stride of their subject
• It seems more likely that we work at a stride by stride level at best
• This analysis would suggest a compression ratio of betwen 50:1 and 500:1, assuming we characterize and specify each stride separately

A different approach is to look at choreographic instructions to ballet dancers, such as represented by Labanotation

• Each action is timed with beats, again at a resolution of about half a second
• This, and similar notations introduce some other concepts which we will discuss in the next chapter.