Identification of promoters is very important in understanding gene regulating relationships in an organism, and computational identification of promoters has been a long standing problem in computational biology. A new method was presented to predict promoter regions in prokaryotic organism. The method predicted transcription unit (TU) first and the TU was divided into singlet that contains only one single gene in a TU, and operon that contains more than one gene. Based on these predicted TUs, promoter was predicted for each TU using hidden Markov model including explicit state duration density. Both predicted TUs and promoters were satisfying.