Partial Matched Filter Bank (PMFB) with orthogonality is a method to sample the phase functions of the received radar signals offering robust Doppler tolerance. Generalized Discrete Fourier Transform (GDFT) with nonlinear phase functions provides engineering flexibility over the traditional Discrete Fourier Transform (DFT). The design freedom of exploiting the entire phase space of constant modulus orthogonal basis brings significant value for the minimization of waveform auto- and cross-correlations that is not possible in DFT. In this paper, we propose the use of GDFT waveforms with optimized correlations in a PMFB framework. It is shown that GDFT based PMFB performs quite promising for Doppler estimation in radar systems.