Feature extraction of ECG signals in meridian systems using wavelet packet transform and clustering algorithms

Chinese traditional medicine (CTM) plays an important role in illness treatment and medicare for a long time due to its safety, effectiveness, low-cost and no obvious side effects. However, the meridian theory, the fundamental theory of CTM, has mainly depended on empirical methods up to now. In this paper, a feature extraction method of meridian system through ECG signals measured at acupoints is presented. The measurement of ECG signals at acupoints was firstly implemented on two meridians of 10 volunteers. Afterwards, the measured ECG signals at acupoints were decomposed by the wavelet packet transform. The decomposed signals show that the energy entropy of ECG signal at an acupoint is larger than that at a non-acupoint. Finally, two clustering algorithms are applied to the partition of energy entropy data at acupoints and non-acupoints, respectively. The obtained clustering results reveal that higher energy entropy of ECG signals at acupoints can be regard as a significant feature to discriminate acupoints from non-acupoints.