Sensing and Adjustment

Sensing and Adjustment is a command and response simulation for training to accurately execute verbal commands in a gunnery scenario.

Currently, training of this variety is accomplished with a trainer checking boxes and listening while the trainee goes through repetitive and routine motions, possibly not meeting the specific command phrase requirements of an actual combat situation. 

The goal of this simulation is to utilize technology where manpower is less reliable and subject to external influences and bias. This allows manpower resources to be better allocated to tasks that require the human component that we do not want in the simulation or combat environments. Additionally, this simulation allows the trainee a unique experience, which is scalable to continually analyze and revisit responses and adapt to multiple environments and scenarios.

Initially, tapped into the native Windows Speech Recognition, which successfully recognized general speech rather than specific phrases. We determined a better solution would need to be built.

Using Python, we analyzed voice data to identify full phrases with accuracy, which we then evaluated those same patterns to be able to differentiate between separate commands. We utilized the Pandas library for data manipulation and processing. We also utilized Matplotlib to graph the data to better visualize and identify patterns that we then turned into the structure of the audio profiler.

Using C#, we took the data from the previous step and built what essentially creates an audio profile for the user, taking sample phrases in the configuration section and comparing them to full phrases in the simulation. 

The end product will ideally give confirmation of whether the simulation requirements were met, possibly on a scored scale, which could be used to present continued learning opportunities to reinforce successful commands and revisit weaknesses. 

Regarding the game as published, additional polishing would be our next steps, such as click here indicators, confirmation of recording, multiple terrains and target sizes, and saved user profiles.

• Troy Galloway, Data Analysis & Programming
  
• Jessica Galloway, Data Testing & Marketing
  
• Eric Gutierrez, UI & Game Design