Searchable Speech for Spacecraft Operations

Spacecraft operations generate voice-loop recordings across separate audio channels. Reviewing those recordings manually made it difficult to find a particular exchange, inspect an error or reuse operational information. Technical vocabulary and varied accents added to the transcription challenge.

My contribution I helped develop an automatic speech-recognition prototype that connected audio capture, transcription, storage and a searchable operator interface. A Raspberry Pi client recorded and transmitted audio; a backend processed the recordings; a dashboard allowed operators to view, search and edit the resulting transcripts.

Engineering focus The recognition system was adapted using domain-specific language data. The interface connected searchable text with the original communications through keyword and date-time references, supporting review and quality-assurance work.

Delivered capability The prototype made recorded voice-loop information accessible as searchable, editable transcripts. It demonstrated the capture-to-review workflow for the client, with deployment beyond the prototype requiring further validation.