Designing an Automated system for Casualty Status Reporting

Designing an automated speech recognition system that allows soldiers to verbalize treatment plans into their headset for the injured and wounded. The system then fills out a casualty report for the medic, streamlining communication and detailing treatments the patient has already received so that medical personnel can quickly create care plans

This research was done for Special Operations Communications in coordinance with MIT Lincoln Labs. The premise of this project is that Automated Speech Rcognition (ASR) program would listen for a wake word such as "soldier down." From there, a soldier would narrate the injuries and treatments that they are performing on their fallen battle buddy. The ASR algorithm would then transcribe the narration, and listen for key words such as "hemorrage" and "tourniquet", and automatically list those treatments and injuries on the Triage Casualty Care Card shown below.

The data would also appear on an app that the onsite medics would have, so that the medics immediately know who is injured, how, and what treatments they have already been administered. This helps speed up the process of soldiers recieving professional medical attention

Goals

• 100% of TCCC Information to reach higher echelons of care

• TCCC Information to reach higher echelons of care before casualty arrives, given a network is in place

• Combat medics freed to use valuable time to fill out TCCC Card

• Collection of battlefield trauma data to enable large military-specific trauma care studies.

Through, this project I am learning how to train and program whisper and NEMO speech recognition models as well as annotate data on label studio

Unfortunately, as this project is through Special Operations, most of the code and other specific information is classified