A Gesture-Based Communication System for Firefighters During Search and Rescue Missions

Fir e departments have primarily utilized radios for communication since the 1980s. Although this method of communication may suffice for non-emergency calls, in the case of a search and rescue, radios have been found to hinder efficient and safe communication. This paper presents a gesture based communication system that reduces the amount of time and effort for communication. There is no standard fire suppressive glove used by firefighters, therefore four gloves were trained with three different predictive models (Random Forrest (RF), k-Nearest Neighbor (KNN), and Long Shoct-Term Memory (LSTM)). LSTM neural network model produced the most accurate results with accuracy percentages between 97.38 % and 98.94 % for all four glove types. Furthermore, a gesture scheme was designed and tested for a search and rescue task, and compared against radio communication. The results showed that the average search time was reduced by 9.30s. Additionally, a survey was conducted by firefighters stating their preference of the gesture system over radio communication and desire to integrate the system into their standard gear.