We propose the use of the Gaussian-curve fitting algorithm for the improvement of measurement accuracy in wavelengthscanned Fabry-Perot filter based demodulation systems. The peak locations of FBG sensors were calculated from the fitted curves rather than from distorted PD profiles, resulting in much better measurement accuracy than that of the highest-peak search algorithm. Also, the algorithm was proved to minimize measurement uncertainty of spectrally-distorted grating sensors. From our experimental results, a temperature resolution as small as ${sim}0.3^{circ}C$ was readily achieved by use of the Gaussian-curve fitting algorithm whereas the highest-peak search algorithm gave a temperature resolution larger than ${sim}4^{circ}C$.