This study aims to propose a novel method to determine the fundamental posture of cycling exercise. The authors have been engaged to design an intelligent system and its control method that can find good saddle height against the physical properties of a cyclist. In order to complete our purpose, firstly, the variation of muscular activity statement due to the difference of saddle height has to be quantitatively evaluated. Pattern entropy method that compresses time-series data with focusing on the instability of data is applied to surface electromyography (SEMG) of lower limbs measured under pedaling exercise. Especially, the condition of upper body was added to the experimental conditions, where a subject grasped a handle and kept hand free in sitting up position during cycling exercise. The pattern entropy method indicated the stability of voluntary muscular activity, however this method is not proper to evaluate the variation of muscular activity statement of lower limb muscle to discuss the cycling exercise.