Modeling of Hysteresis in Human Meridian System with Recurrent Neural Networks

In the theory of the traditional Chinese medicine, it has been found that the acupuncturepoints are distributed in the meridian system of the human body. Moreover, meridian system is an independent system which exists in the body parallel with neural systems and blood circulation systems (Tsuei 1998, Trentini et. al. 2005). The experimental results have shown that the meridian system has significant effect on human health (Tsuei, 1998). Based on the recent research results, it was illustrated that the meridian system had an architecture with many channels allowing the electrical signals passed through easily (Zhang et. al. 1999, Yang 1997). That could be used to explain why the acupuncture-therapy would treat some diseases in human body by implementing some stimulating signals on the related acupuncture points. The acupuncture points distributed in the meridian system possesses some distinctive ways for transferring signals and processing information including electrical information (Yang 1997). Until today, there have been some research results on human meridian system focusing on the analysis of impedance on single acupuncture-point (Yamamoto and Yamamoto 1987, Yang 1997, Zhang et. al. 1999). However, the meridian system is, in fact, a network with several channels. In each channel, there are several acupuncture-points located along a curve. The experimental results demonstrated that there were some relations among those points in each channel. Therefore, the analysis just depending on the impedance of one single acupuncture-point would not reflect the main characteristic of the signal transmission in human meridian system. One of the options is to use an excitation signal to stimulate an acupuncture-point in a channel of the meridian. Then the corresponding response of another acupuncture-point in the same channel is measured. Thus, the signal transmission performance of the measured channel in the meridian can be evaluated. Moreover, the experimental results have demonstrated that the human meridian system is a dynamic system (Zhang et. al. 1999, Yang 1997, Wang et. al. 2009). In this case, the identification of the model to describe the dynamic behavior of the meridian is an efficient way for performance evaluation. Wang et. al. (2009) developed an auto-regressive and moving average model to describe the human meridian system. It fits the response well when the exciting signal with slow frequency and the input amplitude is rather small. However, when the frequency of the exciting input is higher or the amplitude of the exciting signal is larger,