In this paper, an algorithm has been developed to extract the important features of the electrical activity of the stomach measured non invasively with placing electrodes on the abdomen of the human. The measured signal generated from stomach's muscle contraction, is called the Electrogastrogram (EGG), It is a mixture of action potentials with different amplitudes depending on position of the electrodes, direction of spread over stomach, and firing rate. The proposed algorithm is based on special structure of cascaded filters characterized with high selectivity. Parameters of individual section as well as the number of sections were estimated such that minimum mean squared spectral deviation between the measured and estimated EGG signal is achieved. The amplitudes of individual frequencies extracted by this algorithm are considered as features of EGG signal that can be used for studying stomach's physiological states. An example is given to illustrate the application of this algorithm for evaluating the stomach activity during Hunger or Digestion states. The percentage of success to discriminate between these two states was about 93.8 % for the Hunger state, and 98.9 % for the Digestion.