Abstract: ‍ ‍The traditional channel state information fingerprint localization method has the following problems： 1） the channel state information of a single band or single domain has large dimension loss and poor fingerprint interpretation ability； 2） the baseband design of hardware equipment leads to the distortion of channel amplitude and phase， thus leading to poor localization robustness. Therefore， aiming at the current popular dual-band WiFi network card， this paper proposes a dual-band WiFi time-frequency domain joint fingerprint optimization method for indoor localization. Firstly， the amplitude and phase are optimized by obtaining the dual-band frequency-domain channel state information， and then the dual-band time-frequency amplitude joint fingerprint is extracted from the optimized channel state information. The fingerprint of multiple samples is input into the localization model to construct a candidate location set. Then， according to the candidate set， a trustworthy location selection algorithm is proposed， which jointly optimizes the kernel density function and weight of each candidate location， selects the trustworthy location for weighting， and obtains the optimal estimation of the final location. Experimental results in two real-world environments show that the proposed method greatly improves the ability of fingerprint interpretation， and achieves higher positioning accuracy and robustness than traditional methods.