To achieve high-precision spectral retrieval under high-temperature conditions, we compared and analyzed the accuracy and stability of seven different spectral line-shape profiles and developed a spectral line-shape profile selection strategy suitable for high-temperature diagnosis. Firstly, two H2O absorption spectra in the temperature range of 1 100~1 600 K were measured experimentally, and seven different spectral profiles were used to fit them respectively. The integral absorbance, velocity-dependent line-width and Doppler half height and half-width (FWHM) of each measured spectrum were obtained, and the gas temperature value was calculated according to the line-intensity ratio method. The comparative analysis show that under the condition of high temperature and atmospheric pressure, the fitting accuracy of the Gaussian profile is the worst, and the gas temperature inversion accuracy of the Gaussian profile is the highest; Setting the Doppler half-height width parameter in the line-shape profile as a constant can not only effectively improve the stability and accuracy of spectral line parameter inversion, but also improve the temperature inversion accuracy of each spectral line-shape profile. Compared with the non-Voigt profile (speed-dependent Voigt profile, Rautian profile, Speed-dependent Rautian profile and Hartman tran profile), the integral absorbance and velocity-dependent line-width obtained by the Voigt profile are small, and the relative error of calculated temperature are large. Finally, the running time of seven line-shape profile fitting programs was compared. Under the condition of ensuring accuracy, the speed-dependent Voigt profile calculation speed is the fastest. Therefore, the gas temperature is obtained by Gaussian Profile. The Doppler half height and half width are calculated according to the temperature and fixed as a constant, and the speed-dependent Voigt profile is used to fit the measured spectrum, which can effectively improve the inversion accuracy, velocity and stability of high-temperature spectrum.