Most past efforts only focus on haplotyping on genotype data of unrelated individuals and data with general pedigrees but give insufficient effort to the special case of trio genotype data. We present a two-stage method in this paper, which first determines all feasible haplotype configurations for each trio, then uses the expectation-maximization (EM) algorithm to estimate haplotype frequencies based on these haplotype configurations. We design the haplotyping algorithm and the EM algorithm to estimate haplotype frequencies for this special trio genotype data. All algorithms are implemented in a C software package TRIOHAP. Its effectiveness and efficiency have been tested both on simulated and real data sets. The experimental results show that, TRIOHAP runs much faster than the other popular haplotyping software, which discards the trio information. Moreover, because TRIOHAP utilizes such information, our estimation is more accurate and more reliable.