The invention relates to a method and a system for recovering an ideal signal from a distorted signal. A feature of the distorted signal is extracted to serve as an input terminal of an artificial neural network, a feature of the ideal signal is extracted to serve as an output terminal of the artificial neural network, and through training and building the artificial neural network, the accurate feature of the ideal one-dimensional signal is acquired by using strong processing capacity of the artificial neural network to problems with extremely complex environment information, unclear knowledge backgrounds and undefined inference rules. For the situation with variable measurement conditions, the invention also simultaneously designs a computation system, a same computation unit is used repeatedly by only using a parameter update method, and meanwhile requirements on instantaneity and saving of computing resources in the actual system are met simultaneously.