Active noise control by using prediction of time series data with a neural network

To extend applications of an active noise control system, it is required to reduce the distance from the detection microphone to the error microphone. The reduced distance is needed because of the delays of the system. Although it can be changed by hardware improvements, there is a limit to the amount by which the system can be shortened. The purpose of this paper is to develop a predictor that can shorten the distance. We applied a neural network to predict the time series data of the fan noise. Our investigations showed that fan noise had chaotic features and it was possible to predict the near future data of fan noise by a neural network. Applying the predictor using a neural network to active noise control, the prediction improved the performance of active noise control. These results confirm that the prediction technique is useful for active noise control.