<p>In the field of target localization based on Doppler through-wall radar (TWR), accurate estimation of the target’s instantaneous frequency (IF) curve by time–frequency analysis (TFA) techniques is crucial for precise localization. However, in multi-target detection scenarios, existing TFA techniques often suffer from time–frequency aliasing, significantly impacting the estimation accuracy of the target IF curve and thereby limiting improvements in localization accuracy. To address this issue, this paper proposes a target localization algorithm based on generalized fractional Fourier transform (GFRFT). First, a priori knowledge of fractional angles is obtained through the Bézier model to construct a fractional angle dictionary. Next, based on the atomic features of the dictionary, a fractional Gaussian window (FGW) is designed. A multi-dimensional feature fusion method is then employed to integrate time–frequency representations across different fractional domains, effectively suppressing time–frequency aliasing while enhancing the separability of target features. Finally, a time-varying variance estimator with adaptive adjustment capability is designed to dynamically optimize the window width parameter of the FGW, achieving precise matching of the signal’s time-varying characteristics. Based on this framework, the system extracts the IF curve through local energy peak detection, ultimately reconstructing the target’s motion trajectory. Experimental results demonstrate that the proposed GFRFT method significantly improves IF estimation accuracy and localization accuracy by 83.7% and 96.4%, respectively, compared to the STFT method.</p>