Identification of SISO N-L-N Hammerstein-Wiener model is studied. First the model structure is motivated and disturbance model is discussed. Then the work is concentrated on parameter estimation. The original model is approximated using a simpler structure in which error is bilinear-in-parameters so that a relaxation scheme can be used. The relaxation algorithm will determine the input nonlinearity, a high order ARX linear model and the inverse of output nonlinearity. The original model is then obtained by order reduction on the linear model and by calculating the output nonlinearity from its inverse Simulation studies will be used to illustrate the method.