On-line signal trend identification

An artificial neural network, based on the self-organizing map, is proposed for on-line signal trend identification. Trends are categorized at each incoming signal as steady-state, increasing and decreasing, while they are further classified according to characteristics such signal shape and rate of change. Tests with model-generated signals illustrate the ability of the self-organizing map to accurately and reliably perform on-line trend identification in terms of both detection and classification. The proposed methodology has been found robust to the presence of white noise.