The wideband cross-ambiguity function method is commonly adopted to execute the parameter estimation of wideband noise radar, but it needs three-dimensional search in distance, velocity and acceleration when dealing with maneuvering targets, which takes huge computation burden. A novel method based on the conjugate noise group is proposed for addressing the problem of parameter estimation of maneuvering targets. Firstly, the multiple channel is set up according to the echo stretching effect, and the internal signals of the noise group is cut out in fixed length for mixing in each channel. Then the Doppler phase is estimated with the mixed signal by Fractional Fourier Transform (FrFT). The Phase compensation function is constructed by the Doppler phase and the delay is estimated by Frequency-domain Scale Correlation (FSC) algorithm with the compensated noise group signal. Finally, the range, velocity and acceleration are obtained by the two simultaneous equations of the Doppler phase and delay. The proposed method avoids three-dimensional search and reconstruction of the echo signal in time domain, which reduces a large amount of computation compared to the wideband cross-ambiguity function method. The method is feasible for real time processing as the whole algorithm can be accomplished by Fast Fourier Transform (FFT). The effectiveness and superiority of the proposed method are demonstrated by the simulation results.