To solve the  low filtering accuracy problem of the divided difference filter (DDF) algorithm, this paper proposed an support vector regression based adaptive divided difference filter (SVRADDF) algorithm. The difference between the measurement innovation covariance and theory covariance matrix were used as the adaptive factor of the input and output of the support vector regression machine for realtime correction of the DDF noise covariance and the adjustment of the noise covariance matrix according to the actual noise changes, so as to improve the filter precision. Monte Carlo simulation for underwater target bearingonly tracking systems indicates that, with the same initial noise conditions, the proposed SVRADDF algorithm has a better estimation performance and robustness. The accuracy, stability and covergence time are significantly better than the DDF algorithms.

Key words：