Advanced Speech Recognition Technologies

Head-Mounted Displays (HMDs) used with voice input can support hands-free maintenance with technical information being available at all times. Speech recognition is the preferred method for documenting maintenance activities performed, and for quickly accessing technical information required for a specific maintenance task. The maintenance support system will benefit from a multimedia interface with the flight-line technician. The adaptation of multiple noise abatement and speaker voice tracking technologies is proposed in the system together with TERI Research Inc. (TRI) state-of-the-art speech Natural Language Speech (NLS) recognition, synthesis and HMD interfaces. These technologies include: Parabolic noise canceling microphones, Dynamic noise canceling filters, Random noise canceling algorithms, Stationary noise feature set extraction, Word and Context tracking, Speaker Tracking and Beamforming techniques. The combination of these emerging technologies are expected to be very resistant to accuracy degradation caused by both continuous and impulse noises present in military platforms and on flight-lines, as well as in the presence of other voices. An integrated micro-miniature system will be designed that is able to accurately interpret spoken statements and speak responses in dialog fashing, effectively controlling maintenance technical information and displaying it on the HMD. Unique to this research is the approach that will maintain high accuracy speech recognition in the presence of changing and variable high noise environments without having to change the speech corpus or user profiles. The TRI developed state-of-the-art speaker independent speech recognition and synthesis software will perform speech recognition and voice response in real-time, maintaining speech recognition and transcription accuracy in a high noise environment.