Seen the importance of human being language, its treatment became an active research domain, crossing of numeric signal treatment and symbolic language treatment. Since years 60, this discipline benefits very important research, in the two axes recognition and synthesis. The first aims to recognize the user's messages to translate them in action whereas the second has for objective to endow the computer of capacities to read texts.

Language recognizing and synthesizing constitute a challenge to permit passage to the next generation of interfacings man-machine. In spite of the advanced achieved these last years in these domains, progress remain to increase the comfort of present system.

This thesis, that integrates in language treatment setting, titles “ Survey of Arabian language phonetics for recognition “ has for goal to achieve a survey on Arabian language in order to extract, the different phonemes, then diphones that take account of the phenomenon coarticulation and in triphones. These units will be used in recognition motor for continuous Arabian language.