In this paper, we studied a signal processing method which extracts each MUAP(motor unit action potential) from EMG(Electromyogram) interference pattern for clinical diagnostic purposes. First of all, differential digital filtering is selected for eliminating the spike components of the MUAP' s from the backgroud noise. And the algorithm identifies the spikes over the certanin threshold by template matching in frequency domain. After missing or false firing factor is cut off at the IPJ(inter pulse interval) histogram, we average the MUAP waveforms from the raw signal using the identified spikes as triggers, and finally, measure their amplitudes, durations, and numbers of phases. Specially, we introduce algorithm performed by template matching in the frequency domain. This algorithm performed as follows. A typical 3 second signal recorded from the biceps brachii muscle using a conventional needle electrode during a isometric contraction. And decomposed active MUAP' s decomposed from original EMG signal.