Non-invasive recordings of intestinal myoelectrical activity (Electroenterogram, EEnG) are affected by very-low-frequency (VLF) interferences, respiration, ECG and movement artifacts. In order to identify the intestinal pacemaker activity (slow wave), VLF interferences and respiration should be removed from abdominal surface recordings. In this paper a method based on empirical mode decomposition is proposed to identify and cancel such interferences and to enhance external recordings of human EEnG. The study was carried out on 10 recording sessions in which it was acquired 3 surface EEnG signals by means of active laplacian electrodes, together with respiratory and ECG signals. The application of the proposed method permits to increase the signal-to-interference ratio of EEnG signals from - 7.9 ± 3.3 dB to 6.1 ± 2.7 dB. In addition, the method helps to identify the intestinal pacemaker activity from surface EEnG recordings: the dominant frequency of processed signals is 8.6 ± 1.7 cpm which fits the frequency of the intestinal slow wave in the jejunum, whereas the dominant frequency of original signals was 1.6 ± 1.1 cpm which is associated to VLF interferences.