Modeling of psychophysiological processes in the technological framework is helpful in understanding them and also useful in modifying them when their regulation is necessary. In this paper, a new theoretical model of the learning process in biofeedback training is proposed which is closely related to the control theory and the neural network theory. In modeling, biofeedback is regarded as an acquisition technique of self-regulation ability, which is characterized by (1)addition of outer informational pathway to feedback the inner physiological condition to the sense, (2)mental training on the level of consciousness referring the information from the sense, and (3)eventual acquisition of ability to voluntarily control the inner condition without the outer pathway. These three aspects of biofeedback must be included in the model. The model is based on the recent theory of skill learning. For the skill maturation of voluntary control there has been proposed a learning model, in which existence of a self-organizing system to learn a feedback control system referring its input-output characteristics is hypothesized. In our model of biofeedback, we hypothesize applicability of the self-organizing system to feedback control activities in the subconscious level as well as its existence. The model system operates in either of two modes : learning mode and voluntary control mode. At first, it operates in the learning mode, where the self-organizing system organizes itself so that the error correction signal in the target feedback controller vanishes. When the learning process converges, its characteristic shows the inverse characteristic of the regulatory system which has been controlled by the feedback controller. Therfore, a new feed-forward controller of the regulatory system is created through the learning. The model system, then, changes its mode to the voluntary control mode. In the mode, the outer informational pathway is removed. If a voluntary signal to regulate the subconscious activity appears, it drives the newly created feed-forward controller. Since it is the inverse system of the regulatory system of the subconscious activity, it is easily (i.e. voluntarily) controlled following one's intention. In order to evaluate the model performance, it is applied to the first order control system with the learning system structured on the neural network. After some iteration learning converged, and the system became a feed-forward controller which override the target controller and output of the regulatory system smoothly pursued input waveforms. Moreover, at this stage it did not require the explicit information concerning the target controller, which means that it was operable without "outer pathway". Those results show validity and effectiveness of the model for biofeedback. Thus, it is concluded that this model involves three crucial aspects of biofeedback and enables quantitative approach to its learning process.
View full abstract