A novel fuzzy system whose fuzzy membership functions are the reproducing kernel functions is established to detect weak signals from a chaotic clutter by employing the approximating property of the reproducing kernel functions, which can generate the best structure automatically and avoid falling into local minimum with global convergence. Parameter learning is restricted to the antecedent. The proposed fuzzy system is applied to the estimation of the weak signal in chaotic clutter. The results show that the weak signals in chaotic clutter are accurately detected, which demonstrates the effectiveness of the systems.
