BABIR - A better audition for a better interaction with humanoid robot

A robot has to offer an efficient and intuitive interaction with the user in order to be considered as an efficient co-worker in an industrial environment, or as a real companion in a domestic environment. Speaking to the robot and listening to its feedback are the natural interaction way required by potential users.

Speech recognition is a rather mature technology for some dedicated applications, as for example telecommunication or answering machines. In these applications, the user speaks in a dedicated microphone and the speech analysis process gives satisfactory results. In other fields, speech recognition is still disappointing: the surrounding noise and the bad quality of signal acquisition do not allow a robust recognition. The vocal interface is generally given up for the benefit of keyboards or tactile screens.

In this project a robust vocal interface between human and robot will be developed. It will be implemented on Nao, a small humanoid robot. Experiments will be performed in a domestic environment.