Analysis to Two Detecting Methods of Non-Linear Analogue Quantity

In the embedded distribution control system, the analogue quantity under test is usually non-linear. In this paper, the concept of variable threshold neuron for the detecting of non-linear analogue is adopted. The subsection linearization and subsection variable slope are chosen as its training methods. It is shown by analyzing and comparing for the two training methods that the subsection linearization training method can improves the detecting precision and detecting resolution, and the subsection variable slope training method can not only calibrate the detecting curve from the subsection linearization ones but also give less computation error than it. Their simulation results are provided, and precise and feasible measurement is realized.