Applying optical motion capture system has significant effect to rehabilitation since they enable data collection and quantitative assessment of coordinated joint movement. However, introducing the system to assess the training of upper limb prosthesis operation in real time on-site is challenging. The measurement range, accuracy and precision are influenced by the properties of camera-marker relation, especially on the shoulder joint movement. In this research, we applied robust parameter design, an assessment method of system parameter harmonization in Quality Engineering, to quantitatively evaluate and discuss the camera layout focused on upper limb prosthesis activity assessment. Fundamental function was set as actual and measured joint angle relation. Thirteen reflection markers were placed on the body, and shoulder joint flexion and abduction were measured with 4 cameras. The experimental result and analysis show the most robust and efficient condition of 4 camera layout was neighboring sided to the target arm, with 2^nd camera at the height of 0.40m below to the acromion, 3^rd camera at the height of 0.40m above the acromion, and 4^th camera at the height of 0.40m above the acromion. The confirmation test result indicated the acquired setup is reproducible.

Expression functions were additionally developed on T-TAC Caption, a remote captioning system for hearing impaired students, based on the results of a questionnaire survey conducted in advance. Two expression functions are “text is sent to the instructor” and “a taker speaks instead”. An evaluation test was conducted in a simulated class, and the effectiveness of the expression functions was evaluated using Scheffe's ANOVA on paired comparison (Nakaya variation). There was a significant difference in case of “Comprehensive” viewpoint combining the three viewpoints of “Psychological burden”, “Accuracy”, “Immediacy”. Students that were hearing impaired were found to make decisions by comprehensively taking all viewpoints into account.