Image training system for archery: Hatayama 2005

The authors built a motor learning system which, as well as assessing the form and timing of a movement, lets the learner record their own sensations. On a three-dimensional avatar the learner marks spheres at the point and the moment where they felt something, for instance tension in a shoulder, and marks in the same way the mental image of the movement they are aiming at; they can then review their own movement alongside the spheres of an earlier one and compare the two. The test case chosen is kyudo, because the movement is prescribed in its eight stages and does not change from one trial to the next.

The trial was run on five people with no experience of kyudo, split into two groups: two used only the part of the system that assesses the movement and plays it back, three also used the recording of sensations and the comparison with earlier trials. Fifteen trials each. The form score computed by the system, out of 100, rose in all five, but the mean rise from the first trial to the last differed by 25 points between the two groups, in favour of those recording their sensations. The number of spheres marked grew over the first trials and fell towards the end: the authors read the growth as the ability to attend to several places at once and the fall as the movement starting to come without conscious attention.

The other result is a problem. For the same number of trials, those using the full system took longer, because recording sensations takes time, and because an image once marked had to be marked again in order to come back to it on the following trial. In the closing questionnaire, filled in by the three who used the full system, the answers to the six questions fall between 4.0 and 4.7 on a five-point scale; in their free comments the participants say it is hard to mark at the right moment while reviewing the movement, and that spheres alone are not enough to represent what they felt.

The movement was filmed with a USB camera and image-processing software that locates objects of a chosen colour: markers of different colours were fixed to the learner's joints, and the coordinates in the image plane were converted into three dimensions with the DLT method. For the parts that move faster and demand more precision, that is the head, shoulders and hands, an optical HiBall-3100 sensor and an InterTrax2 sensor were used instead. The points measured are the head, shoulders, elbows, hands, both sides of the hips, the toes and the heels.

The assessment of form rests on the norms of kyudo. The sanju-jumonji requires the line of the shoulders, the line of the hips and the line joining the toes to be parallel and in the same plane, with the centre line of the chest crossing them at ninety degrees; the goju-jumonji requires further right-angled crossings between bow and arrow, bow and left hand, right thumb and string, chest line and shoulders, and the hollow of the neck and the arrow. Of these crossings the system measures six: the one between neck and arrow is left out because, according to the federation manual, the correct orientation of the face differs from archer to archer, and the crossing between chest and shoulders is counted once. The score is worked out by deduction, out of 100 points for each of the eight stages, measuring every 10 milliseconds the angular deviations of the crossings and the positional deviations of the shoulders, elbows and hands from a reference movement recorded beforehand, with a correction for body size. Timing is assessed separately, comparing with the reference the moment the learner starts to move and the moment they stop, judged from the coordinates of the two hands, and displayed as two superimposed charts.

Five people in all, two against three, with no randomisation stated and nobody blinded: the 25-point difference is a mean of gains, given with no spread, no test and no probability value. The score that measures the outcome is produced by the same system that does the teaching, and by deduction from a reference movement recorded in advance: it is not an outside measure, and improving on it means moving closer to that reference. The two groups did not even work for the same length of time, since the full system demanded more of it for the same number of trials.

Nobody shoots at a target. There are no hits, no scores, no distances: the study is about the form of the movement in the early stages of the hassetsu, in people who had never held a bow, and it says nothing about what happens once aiming begins. The questionnaire covers three people. The paper does not report the accuracy of the tracking, which here comes from a colour camera, an optical sensor and an inertial sensor combined, and there is no follow-up. It is a piece of engineering work from 2005, written to present a system: the trial on five learners is there to show that the system works, not to establish the worth of that way of teaching.

For an archer today the interesting result is not the number but the idea. A learner working on a prescribed movement usually receives feedback on form and timing, that is on what can be seen from outside; here the attempt is to keep track of what they feel as well, and to put it back beside the form so that good trials can be compared with poor ones. It is the intention behind a training diary and the voice notes taken after a session, carried into a tool that shows them at the point and the moment where they were felt.

What the study can support is little: five kyudo beginners, fifteen trials each, a form score computed by the same system, and those recording their sensations improved more. There is nothing about the target, nothing about Olympic recurve shooting and nothing that outlasts the session. What remains is the authors' observation about the number of spheres, which first grows and then falls, because that is the kind of pattern an archer recognises when a movement stops being followed piece by piece and starts to come on its own; in the study, though, that fall is described in three people and is not related to any shooting result.