Speech Translation Apparatus Automatic Speaker Direction Detection

The system automatically identifies speakers and switches translation directions without user intervention. The sound source direction estimator and controller work autonomously to detect which user is speaking and adjust the translation output accordingly, eliminating the need for manual button operations and achieving self-service operation.

The patent replaces manual mechanical button operations with an automated acoustic field-based identification system. Instead of users physically pressing buttons to switch languages, the system uses microphone arrays and sound source direction estimation to automatically detect speaker identity and switch translation directions programmatically.

The patent introduces a sound source direction estimator as an intermediary between the acoustic signals and the translation system. This intermediary component processes the raw acoustic signals from the microphone array, estimates sound source directions, and provides cleaned speaker identification information to the controller, filtering out noise interference in the process.

The system uses multiple microphones in an array to create multiple copies of the acoustic signal from different spatial perspectives. By comparing these copied signals and their directional information, the system can accurately identify the sound source direction even in noisy environments, as the spatial redundancy helps filter out unrelated noise.