The passive voice often appears in patent documents but seldom gets the right translation results,which has greatly affected the understanding of the full text.This paper focuses on Chinese analysis and English generation of passive sentences in patent documents,explores the possibility of transforming the sentences into passive voice from semantic perspective based on hierarchical network of concept (HNC) theory,analyzes the characteristics of passive voice both in Chinese and English documents,summarizes three states of the sentence when translating Chinese into English in processing the voice:keep the active voice,passive sentence with grammar mark in Chinese,passive sentence without grammar mark in Chinese,puts forward a series of methods to resolve the latter,gives the transformation rules and algorithm.Experiments indicate that these methods produce better translation quality and have a bright future in an online patent machine translation (MT) system.