A constrained least square and trimmed least square method for multisensor data fusion

Though neural data fusion algorithms based on a linearly constrained least square (LCLS) method solve the ill-conditioned and singular matrix problems that arise in the LCLS method, they don't perform well when there are impulsive noises attached to several sensors. In this paper, a data fusion algorithm based on a constrained least squares (LS) and trimmed least squares (TLS) method is proposed. On one hand, it inherits the unbiased statistical property and the merit that no priori knowledge about the noise covariance is needed. On the other hand, it is more robust and has better results than LCLS and linearly constrained trimmed least squares (LCTLS).