A Kalman filter approach to virtual sensing for active noise control

Abstract

Local active noise control systems aim to produce zones of quiet at a number of desired locations within a sound field, such as the ears of an observer. The resulting zones of quiet are usually centred at the error sensors, and are often too small to extend from the error sensors to the observer's ears. To overcome these problems, virtual sensing methods have been suggested. These methods are based on estimating the error signals at a number of locations remote from the physical locations of the error sensors. By minimising the estimated error signals, the zones of quiet can be moved away from the error sensors to the locations where noise control is desired, i.e. the virtual locations. In this paper, the active noise control problem under consideration is analysed using a state-space model of the plant. Kalman filtering theory is then used to develop a virtual sensing algorithm that computes optimal estimates of the error signals at the virtual locations. The developed algorithm is implemented on an acoustic duct arrangement, and the real-time estimation performance at a virtual location inside the acoustic duct is analysed. Furthermore, the developed algorithm is combined with the filtered-x LMS, and the results of real-time broadband feedforward control experiments at the virtual location are presented.