More and more linguistic information has been employed to improve the performance of machine translation, such as part of speech, syntactic structures, discourse contexts, and so on. However, conventional approaches typically ignore the key information beyond the text such as prosody. In this paper, we exploit and employ three prosodic features: pronunciation (phonetic alphabet and tone), prosodic boundaries and emphasis. Based on the annotated data, a conditional random fields (CRF) sequential tagger is used to label the prosodic tags for Chinese sentences, and three methods are presented to integrate these features: (1) factored translation models where the prosodic features are incorporated as factors; (2) a word lattice decoding model where the prosodic boundaries are considered to be an alternative to the tokenization boundaries; (3) re-ranking models where the prosodic features are integrated in the language model to re-score the n-best translation candidates. We evaluate the proposed methods with bilingual evaluation understudy (BLEU) score both in English-to-Chinese (E2C) and Chinese-to-English (C2E) translation directions. Experiments show that with prosodic features, the re-ranking model achieves significant improvement, while the word lattice decoding and the factored translation models also improve the performance.