This paper adresses the problem of symbol estimation in the case of a spread spectrum based system where the receiver has only the information about the nonlinear function used to generate the spreading sequence. We propose two methods to achieve this, first based on the already considered DUAL form and a correlator second based one. For the DUAL form we consider multiple Kalman filter implementations suited to the case of nonlinear estimation methods among them Unscented Kalman Filter (UKF) and Exact Polynomial Kalman Filter (ExPKF). Finaly to provide a perfomance evaluation on the proposed methods we obtain throughout Monte-Carlo simulations the Bit Error Rate (BER) characteristics with respect to Signal to Noise Ratio (SNR).