Acoustic Noise Identification Using Fuzzy Modeling Techniques

This paper presents a new approach to acoustic noise identification, by introducing fuzzy modeling techniques. Fuzzy identification is compared to conventional linear identification techniques, using as the system to model a new device called ElectroMechanical Film (EMF), developed by VTT [1]. This device can be used either as an acoustic sensor or actuator. The obtained model represents the behavior of the EMF in a three dimensional enclosure. Practical implementations have several nonlinearities due to the sound field itself, actuators behavior, or the electronic equipment. In order to cope with these nonlinearities, high order linear models are usually identified. However, the implementation of control algorithms in real-time requires simple models due to the computational burden. This fact led to the application of fuzzy identification techniques, particularly modeling and identification based on product-space fuzzy clustering [2]. Fuzzy modeling proved to be accurate for complex and partly known systems, and it can represent highly nonlinear systems in an effective way due to their function approximation properties. The proposed identification technique is applied to real-time noise data, and an accurate model is identified. The obtained fuzzy model is more accurate than a linear model, which is also identified. Moreover, the fuzzy model uses less computational resources, which is important for its future application in active noise control.