Estimation of Arm Curl Repetitions Based on Personal Ability and Condition Using Fuzzy Inference

Resistance training is essential for maintaining health. However, some people do not know how to conduct training, and others perform incorrect training, leading to insufficient effects and injuries. Although there are some systems that propose training, there is still a need for a system that proposes accurate exercise intensity based on individual abilities and conditions to provide more effective training and improved safety. We developed a system to estimate individual repetitions using fuzzy inference. We adopted the Rating of Perceived Fatigue (RPF), Physical Activity Level(PAL), 1 Repetition Maximum(1RM). Experiments were conducted on 15 adult males who do not train regularly. The estimated repetitions count by the proposed system was compared that recommended repetitions assessed by surface electromyography (sEMG) muscle fatigue estimation. As a result, comparing the recommended repetition count (mean=11.93, standard deviation=3.58) and estimated repetition count (mean=11.53, standard deviation=3.49), the close alignment between the distributions of recommended and estimated repetitions suggests that the proposed system adequately adapts to the high variability in individual subject characteristics. Furthermore, the system had a Mean Absolute Error (MAE) of 3.2, and with nine subjects maintaining errors within 3 repetitions, the results indicate that the system can predict recommended values with minimal deviation, demonstrating its effectiveness. However, 4 subjects showed errors of 5 or more repetitions. Issues were identified such as poor accuracy for subjects with extremely high or low 1RM and PAL, and low correlation between RPF and repetition counts. For these issues, future studies must improve the design method of membership functions and introduce new input parameters.