We present a novel speech processing strategy for cochlear implant (CI) based on active nonlinear characteristics of biological cochlea that contribute to hearing under noisy conditions. A simple dual path nonlinear model (SDPN) was developed to utilize the advantage of leveldependent frequency response characteristics in robust formant representation. The model was motivated from the function of basilar membrane so that leveldependent frequency response can be reproduced. Compared to dual resonance nonlinear model (DRNL) which has been proposed as an active nonlinear model of basilar membrane, the proposed model is much simpler and thus better suited to be incorporated in CI speech processor. We developed a CI speech processing strategy based on the SDPN for frequency decomposition and an adaptive envelope detector. It was verified from spectral analysis that the formants of speech is robustly represented after the frequency decomposition by the array of SDPN, compared to a linear bandpass filter array of conventional strategies. From acoustic simulation and hearing experiments on normal subjects, it was verified that the proposed strategy yielded enhanced syllable recognition under speechshaped noise than the conventional strategy.