The NA genes of 395 strains of human H3N2 influenza virus isolated from 1996 to 2005 in China were sequenced, analyzed with bioinformatics tools. The NA nucleotide sequence of phylogenetic tree showed a main evolution branch with multiple short side branches. The strains in the same year may be divided into several branches. There was an obvious lag between vaccine strains recommended by WHO and the Chinese circulating strains in phylogenetic tree of the NA nucleotide. The result also showed no amino acid deletion and insertion in the NA. In NA antigen sites, where including residues 197-199 aa, 431-434 aa and 339-347aa the mutation was higher, in contrast, the residues including 153 aa, 328-336 aa, 367-370aa and 400-403 aa, the mutation was lower. Besides the antigenic determinant sites, there also had the other amino acid mutated highly, such as 18, 23, 30, 93, 143, 208, 216, 221, 249, 265, 267, 307, 385 and 437 aa, among them 143 and 267 mutation were higher than that in antigenic determinant sites, their biological significance are not clear yet. The neuraminidase active-site residues in NA were highly conservative and the same were the disulphide bond and the glycosylation sites in NA. In conclusion, our analysis provides some information for influenza prevention and control and the NA inhibitor medicine application.