Virtual sensing at low computational cost for active noise control

In active noise control, global noise attenuation is not achievable when the size of the controlled environment is significantly larger than the wave length of the noise, and sat- isfactory performance is obtained only in the immediate vicinity of the sensors. There is therefore need for methods that can extend the control to positions that are different from the sensor locations, such as virtual sensing. The latter, unfortunately, requires a model to determine the control, which entails a high computational cost if the system dynamics are identified fromdata. This paper deals with the problemof reducing the computational cost required for the system identification of a one dimensional enclosure. Beginning with an analytical model for virtual sensing, we here propose to employ two microphones instead of one as in the usual virtual sensor schemes. This is beneficial in that it greatly reduces the complexity of the involved model. A system identification procedure is then used to obtain the virtual sensor. Both techniques (the analytical and the identification-based) are finally evaluated, by comparing the estimated signal to the measured one.