Based on continuous spectrum analysis, the mathematical model for spectrum signal was established. And the spectrum signal’s systematic error processing method based on the invariance of the ratio of the light intensities at any two wavelengths in the range of continuous spectrum was put forward. Combined with wavelet multi-resolution filtering noise processing techniques, the background interference processing method was established based on the spectral characteristics of the measured water quality parameter. These signal processing methods were applied to our independently developed multi-parameter water quality monitoring instrument to on-line measure COD (chemical oxygen demand), six valence chromium and anionic surfactant in the normative and actual environmental water samples, and the monitoring instrument had good repeatability (10%) and high accuracy (±10%) to meet the technical requirements of national environmental protection standards, which was verified by the contrast experiment with China national standard analysis method for determination of the three water quality parameter. The results showed that the researched signal processing methods were able to effectively reduce the spectrum signal’s systematic error and the interference from noise and background, which was very important to improve the water quality monitoring instrument’s technical function.