This paper considers system identification based on mixed binary-valued and precise measurements with noises. The optimal Quasi-Convex Combination Estimator (QCCE) is constructed by combining the identification results of binary-valued and precise sub...
This paper considers system identification based on mixed binary-valued and precise measurements with noises. The optimal Quasi-Convex Combination Estimator (QCCE) is constructed by combining the identification results of binary-valued and precise subsystems. The convergence properties are theoretically analysed in terms of strong consistency and asymptotical efficiency. Compared with the maximum likelihood (ML) estimator based on the whole system, the optimal QCCE has a smaller asymptotical variance that can achieve the cram´er-Rao (CR) lower bound. Extensive numerical simulations validate the superiority of the optimal QCCE.