Extracting soil nutrient information using spectral analysis technology is very important for precision crop management. Difference indices (DIs), normalized indices (NIs) and ratio indices (RIs) were calculated from all available combinations with reflectance of two random bands between 1 000 and 2 500 nm, and correlated to soil organic matter (SOM), total nitrogen (TN) and available nitrogen (AN) content using dried soil samples from five different soil types in central and eastern China, then key spectral index and quantitative models for estimating SOM, TN and AN content were developed. Results show that spectral indices composed of reflectance corrected with multiplicative scatter correction (MSC) and collated with Savitzky-Golay (SG) smoothing method were highly correlated with the three soil nutrient indices. The sensitive wavebands well correlated simultaneously with the three nutrient indices were 1 879~1 890 and 2 050~2 100 nm, and DI(CR_{1 881}, CR_{2 070}) composed of the reflectance in wavebands 1 881 nm and 2 070 nm corrected with MSC and SG smoothing methods performed the best. Testing of the estimating models based on DI(CR_{1 881}, CR_{2 070}) with independent datasets from different types of soil samples resulted in validation of R², RMSE and RPD being 0.83, 0.79 and 0.72, 0.20 g kg^-1, 4.71 g kg^-1 and 23.96 mg kg^-1 and 2.56, 2.30 and 2.93 for TN, OM and AN, respectively, which indicates that DI(CR_{1 881}, CR_{2 070}) might be a good spectral index for estimating contents of the three nutrients simultaneously.