抄録
In the medical field, the measure against noise is one of the most important problems for measurements of signals from the human body. By the existing methods, the artifacts can not be reduced easily. Therefore we have tried to eliminating the artifacts. In this paper, the objective signal is ECG(electrocardiogram). Our purpose is to eliminate the artifact to extract ECG in the input signal which is comprised in EMG (electromyogram). We have adopted the perceptron model with learning for solving this artifact problem. This model which we have designed is able to learn by referring to the target signal. In this paper the first priority is the time for learning. It is the major premise that the error between the target signal and the output signal is minimized. To minimize the time for learning we have kept this model in small parts. This model we have designed is applied to two simulations and one practical problem. As a result of simulation, we have found this model can eliminate about -20dB noise. The results of experiments showed the applicability to the practical use.
