Similarity Index Values in Fuzzy Logic and the Support Vector Machine Method Applied to the Identification of Changes in Movement Patterns During Biceps-Curl Weight-Lifting Exercise.

Background/Objectives: Correct supervision during the performance of resistance exercises is imperative to the correct execution of these exercises. This study presents a proposal for the use of Morisita-Horn similarity indices in modelling with machine learning methods to identify changes in positional sequence patterns during the biceps-curl weight-lifting exercise with a barbell. The models used are based on the fuzzy logic (FL) and support vector machine (SVM) methods. Methods: Ten male volunteers (age: 26 ± 4.9 years, height: 177 ± 8.0 cm, body weight: 86 ± 16 kg) performed a standing barbell bicep curl with additional weights. A smartphone was used to record their movements in the sagittal plane, providing information about joint positions and changes in the sequential position of the bar during each lifting attempt. Maximum absolute deviations of movement amplitudes were calculated for each execution. Results: A variance analysis revealed significant deviations (p < 0.002) in vertical displacement between the standard execution and execution with a load of 50% of the subject's body weight. Experts with over thirty years of experience in resistance-exercise evaluation evaluated the exercises, and their results showed an agreement of over 70% with the results of the ANOVA. The similarity indices, absolute deviations, and expert evaluations were used for modelling in both the FL system and the SVM. The root mean square error and R-squared results for the FL system (R² = 0.92, r = 0.96) were superior to those of the SVM (R² = 0.81, r = 0.79). Conclusions: The use of FL in modelling emerges as a promising approach with which to support the assessment of movement patterns. Its applications range from automated detection of errors in exercise execution to enhancing motor performance in athletes.