The new method treated the problem of parameter estimation and filtering for chaotic time series as a nonlinear minimisation process and solved it by using a steepest descent algorithm.