The purpose of this study is to estimate the endurance threshold in terms of muscular fatigue during bicycle ergometer exercise. The problems to be solved are induced by dynamic movement and the physiological variation of muscle activity: that is, the progression and impairment of muscle activity occur simultaneously. First of all, we used multichannel recordings of myoelectric (ME) signals to reduce the effect by the movement of a bipolar surface electrode relative to the innervation zones. Second, since even the different types of ME parameters contain redundant information on muscular fatigue, we used the principal component analysis (PCA) to represent the meaningful information by small dimensions. Moreover, we proposed a total evaluation pattern to discriminate muscular fatigue from progression of muscle force at a glance. The total evaluation pattern shows the proportion of first principal component, the components of the first eigenvector, and the correlation coefficients as a function of the work load. The assessment using the total evaluation pattern divided eight subjects into three groups, whereas these subjects were not identified by a specific ME parameter.