A data-compression algorithm for digital Holter recording using artificial neural networks (ANNs) is described. A three-layer ANN that has a hidden layer with a few units is used to extract features of the ECG (electrocardiogram) waveform as a function of the activation levels of the hidden layer units. The number of output and input units is the same. The backpropagation algorithm is used for learning. The network is tuned with supervised signals that are the same as the input signals. One network (network 1) is used for data compression and another (network 2) is used for learning with current signals. Once the network is tuned, the common waveform features are encoded by the interconnecting weights of the network. The activation levels of the hidden units then express the respective features of the waveforms for each consecutive heartbeat.

No abstract

We propose a potential eld approach to represent a game situation. In a potential eld, a ball should be moved according to the gradient of the potential eld. There are three kinds of potential elds. One is dened for a game eld, and another is dened for each robot. The third eld is dened as the combination of these two. The combined eld is used for evaluation of a situation. We applied this method to our robot control program. Potential values are used to determine the direction in which a robot kicks a ball. We compared the potential eld based strategy and an usual \if then" type rule based strategy. The potential eld based strategy makes better decisions in several cases and no worse decisions than the rule based strategy.

Two key features of successful multi-robot systems in RoboCup are robustness of a vision system and optimization of feedback control in order to reach the goal point and generate the action under the team strategy. This paper proposes two new methods. One is a fast image processing method, which is coped with the spatial variance of color parameters in the field, to extract the positions of robots and ball in 1/30 sec. The separation problem in the interlaced format image is solved. Another one is a path generation method in which the robot approaches the goal by changing its direction convergently. With these two algorithms, the real time processing system is realized by generating a stable path under a low quality input image.

No abstract