Indoor cycling was commonly examined from the riding posture, saddle height or pedal force to analyze the muscular activity on cyclists' lower limbs. While strong muscular strength and proper riding posture are important to minimize strain, the significances of these attributes on cycling fatigue were unclear. An attempt was made to identify significant attributing features for indoor cycling fatigue classification based on an experimental study involving twenty healthy postgraduates. The participants were tasked to perform an indoor cycling fatigue experiment at 6km/h with gradual speed increment till fatigue level achieved. The accelerometry, sacral trajectory and the lower limb kinematic changes were measured. Significant feature subset selection was determined using the wrapper approach with IBk algorithm. The featured data were later classified on IBk, SMO, ZeroR, J48 and Vote followed by subsequent discriminant analysis. The results demonstrated that the significant attributes yielded 95.0% and 75% classification accuracies (training data) but yielded 72.5% and 65.0% (10 folds cross-validation) on Vote and the discriminant analysis respectively. Findings revealed that the cycle frequency is the most significant attribute exerting a major effect on cycling fatigue while StepRegML and Disp_ML attributes contribute little to distinguish fatigue and pre-fatigue cycling motion.
展开▼
