Artificial neural network for additive noise filtering techniques

This paper proposes an artificial neural network approach to filter additive noise in time-domain. A backpropagation algorithm is applied and a delta learning rule is employed to update the weights during the training. Signal estimation is tested by a feedforward neural network after the training. The signals which consist of a sinusoid with different levels of white noise are tested for evaluating the performance of the network. Also, the performance of the neural network is compared with a linear-prediction filter. Preliminary test results show acceptable performance