摘要 |
PURPOSE:To realize a real time voice translation by generally processing an anticipated translating operation at the time of inputting the voice before the voice is inputted. CONSTITUTION:Before the input of the voice, a network stored in a recognition network memory part 1 is converted into a translation and voice synthesizing network in a network converting part 2 for performing a translation processing and a phonetic processing and stored in a memory part 4. At this time, a converting matrix is formed in the converting part 2 and stored in a memory part 4. At the time of inputting the voice, the voice is recognized in a successive voice recognition part 5 and a bus on the recognition network through which the voice passes is outputted. In a matrix arithmetic part 6, a translation result is obtained from the converting matrix of the memory part 4 and the recognition result of the recognition part 5. In a search part 7, the voice synthesizing network is searched by the use of the translation result obtained in the arithmetic part 6, and a phonetic sign corresponding to the translation result is obtained to synthesize the voice in a successive voice synthesizing part 8. |