Motion estimation framework and authoring tools based on MYOs and Bayesian probability

Nowadays, diverse kinds of user interfaces are being developed based on the natural user interface/experience. Examples of these include Leap motion, which measures finger motions to produce finger-based commands and MYO, which measures arm motions for arm-based commands. However, these types of motion sensors are still too expensive to be utilized for commercial applications. Moreover, multiple motion sensors sometimes need to be utilized concurrently in order to estimate user motions accurately. Thus, either the cost of motion sensors or the number utilized needs to be reduced. This paper proposes a motion framework that estimates unmeasured motions based on Bayesian probability and measured motions, where motions are defined by a set of MYO sensor values. Bayesian probability is calculated in advance by measuring co-related motions and counting the occurrence of these measured co-related motions. As a result, the number of MYOs needed is reduced. In experiments conducted using MYOs, the processes used to calculate Bayesian probability and to estimate unmeasured motions were validated. Comparison of the measured motions with the unmeasured motions showed that the difference between the two types of motions was small, and indicated that the proposed motion estimation framework estimates unmeasured motions with an average error of 0.05, which exhibits a 25 % improvement over the traditional method.