Abstract: MISP method as proposed by E,Todini in 1977 is currently used to estimate aarameters and states simultaneously with a deficiency of piori knowladge. This method has an inherent slow rate of convergence and a trend of divergence. This paper analyzes such convergence problem and points out that unsuitable formulas are used in adapting measurement noise covariance matrix in the original MISP method. The paper presents a method based on an innovation to accelerate the rate of convergence and proposes a new formala to estimate the measunement noise covariance matrix. Results of simulations with artifical pnd actual data show that the modified MISP method has overcome the couverg-euce difficulties in the original MISP method aud achieved a much higher rate of couvergeuce