Myoelectrical recordings of the gut contains slow waves (slow rhythmicity) and spikes (fast rhythmicity). While the slow wave determines the frequency and propagation of gastrointestinal contractions, spike activities are directly associated with the contractions. Traditionally, spikes are extracted from the myoelectrical recording using high-pass/bandpass filters. Due to sharp rising edge (high-frequency component) of the slow wave, the conventional method is not accurate in the separation of the slow wave and spikes, although it has been used for years. In this paper, a novel and fast blind source separation algorithm was developed. Experimental results showed that the proposed method was able to accurately extract spike activities from the myoelectrical recordings obtained in dogs and that its performance was not affected by the high-frequency components of the slow wave.