Electromyogram (EMG) signals have the potential to allow patients without total muscle controls to operate electronic or mechanical devices, such as a power wheelchair. A muscle contraction can be interpreted as a switching signal and its strength can determine the intended level. The purpose of this study is to implement an algorithm to accurately detect the strength of a muscle contraction. Muscle contraction strength can be increased by temporal or spatial recruitment. In this case, the graded muscle contraction is detected by the temporal recruitment, where the frequency of the EMG signal is evaluated. A nonlinear detection algorithm is used to define the duration of a contraction episode. The frequency of the spikes within the contraction episode is used to determine the contraction strength. The algorithm should be useful for designing myoelectrically controlled devices.