Takayuki Ogasawara, Hanako Fukamachi, Kenryu Aoyagi, Shiro Kumano, Hiroyoshi Togo & Koichiro Oka

IEEE Trans. Human-Machine Systems (THMS), vol. 51, no. 3, pp. 221-228, 2021.

A key skill in archery is the ability to suppress postural tremor while aiming at a target. Providing feedback during daily archery practice is a potentially effective way of suppressing tremor. However, postural tremor is subtle and difficult to measure using vision-based techniques. This article proposes a feedback method that uses a bow equipped with a small, lightweight acceleration sensor. First, we automatically detect an archer’s shooting execution cycle, including the aiming, release, and follow-through phases, by using binary classification, and then, we quantify postural tremor during aiming. Then, from the quantified postural tremor, we regress the expected total score that the archer would obtain in a series of shots during a real game. We performed an experiment with 11 members of a university archery club and achieved 1) a precision of 0.72 and recall of 0.80 in shooting detection and 2) an absolute correlation coefficient of 0.74 in score prediction with leave-one-subject-out cross-validation.