An increase in neuromuscular activity, measured by electromyography (EMG), is usually observed during vibration exercise. The underlying mechanisms are however unclear, limiting the possibilities to introduce and exploit vibration training in rehabilitation programs. In this study, a new training device is used to perform vibration exercise at varying frequency and force, therefore enabling the analysis of the relationship between vibration frequency and muscle fatigue. Fatigue is estimated by maximum voluntary contraction measurement, as well as by EMG mean-frequency and conduction-velocity analysis. Seven volunteers performed five isometric contractions of the biceps brachii with a load consisting of a baseline of 80% of their maximum voluntary contraction (MVC), with no vibration and with a superimposed 20, 30, 40, and 50 Hz vibrational force of 40 N. Myoelectric and mechanical fatigue were estimated by EMG analysis and by assessment of the MVC decay, respectively. A dedicated motion artifact canceler, making use of accelerometry, is proposed to enable accurate EMG analysis. Use of this canceler leads to better interpolation of myoelectric fatigue trends and to better correlation between mechanical and myoelectric fatigue. In general, our results suggest vibration at 30 Hz to be the most fatiguing exercise. These results contribute to the analysis of vibration exercise and motivate further research aiming at improved training protocols.