The method of differential optical absorption spectroscopy (DOAS) is applied to the retrieval of ammonia concentration. Genetic algorithm, wavelet transform and Kalman filtering are all studied to show the inversion accuracy of low-concentration ammonia based on the traditional DOAS algorithm. The results show that all the three algorithms have good denoise performance for ammonia absorption spectrum with low SNR. In the case of ammonia concentration lower than 7.6 mg/m3, the inversion accuracy for three improved algorithms is obviously improved compared with the traditional algorithm. Standard deviation and peak to peak value of the differential absorbance residuals based on Kalman filtering combining traditional polynomial fitting are 0.1349% and 0.6132%, respectively, which shows the best performace. The measurement error is within ±0.3%.