The generalized hidden Markov model (GHMM) is an important model for computational gene finding. Compared with the traditional hidden Markov model (HMM), GHMM needn't the assumption that the length of each state is geometrical distribution, while it is necessary for HMM. This property is appropriate for computational gene finding. The demerit of GHMM is its high computational complexity, which hinders it from being used practically. According to the characteristic of gene's structure, a novel simplified algorithm is proposed without any additional assumptions, and its computational complexity is linear with the length of sequence. The testing result for biological data demonstrates that the simplified algorithm is effective.