Optical fiber temperature sensors based on fluorescence lifetime are widely utilized in a multitude of fields. The integral ratio method (IRM) and fast Fourier transform (FFT) method are the most commonly employed techniques for obtaining fluorescence lifetime. However, both methods have been found to have insufficient precision to varying degrees. We employed the extended Kalman filter (EKF) algorithm to fit the fluorescence attenuation curve, and the weighted linear least squares method is utilized to obtain a more accurate fluorescence lifetime. A comprehensive hardware system is fabricated for verification purposes. The software simulation demonstrates that the EKF algorithm enhances the precision of the fluorescence attenuation curve by up to 96.469%. The results of the test demonstrate that the temperature sensor we proposed has a resolution of 0.07°C. In comparison to the IRM and FFT methods, we enhanced the average accuracy of the fluorescence lifetime by 17.819% and 4.626%, respectively, whereas the average deviation of the temperature is reduced by 80.224% and 54.701%, respectively. We provide an effective way for the algorithm design of high-accuracy optical fiber temperature sensors based on fluorescent lifetime.