Application of an ANFIS Algorithm to Sensor Data Processing

Smart sensors are becoming very popular; one of their main advantages is the capability to perform some extra processing on raw data. In this way it is possible to overcome intrinsic limitations due to transducing technique; for example, it is possible to use low very simple sensors and apply linearization techniques to improve performances without increasing the cost. Obviously, the adopted algorithm should be as simple as possible. In this paper, a new calibration technique is presented; it is based on adaptive neuro-fuzzy networks and it has shown superior performances with respect to traditional algorithms, despite its easiness of implementation. Feasibility has been demonstrated with a pyroelectric biaxial positioning system, reaching the limit of the adopted test bench, in the order of 20mum