Feedforward IIR active noise control using genetic algorithm

Presents an active noise control (ANC) algorithm using a genetic algorithm with infinite impulse response (IIR) filter structure. Stochastic gradient algorithms such as the least mean square (LMS) algorithm are conventionally used for their simplicity and stability. But these algorithms have disadvantages of local minimum and large eigenvalue disparity of input signal´s autocorrelation matrix. To solve those problems, several researches have been made into adaptive filtering schemes based on genetic algorithms. But their application is limited to the system that is able to know the desired signal directly. In general active noise and vibration control systems are not able to sense the desired signal directly, so they have some difficulties or problems such that learning sample set is proportional to population size. The proposed active controller is composed of genetic controllers that can learn by one sample set per one generation. This structure can be properly applied to active control systems. Computer simulations show that the proposed genetic structured IIR active controller has a more optimal result than feedforward active noise control systems