Voice_controlled_drawing_interface

A deep neural network that recognizes voice commands embedded into a small user interface.

Demo of the app

demo.mp4

Run the app

To run the program just download the "model.h5" and "voice_command_app.py" in the same directory.
Navigate to that directory with CMD (for windows or equivalent for your OS) and run the program with the command:
"python voice_command_app.py"

App possibilities

There are 5 default controlls ['izbrisi', 'krug', 'kvadrat', 'oboji', 'trougao'](in serbian) or 'clear', 'circle', 'square', 'color', 'triangle'.
The circle, square, triangle commands draw the corresponding shapes. The color command colors the corresponding shapes in a random color.
The clear command clears the screen. There is a "Voice command" button which records input from the microphone for 2s, and then the recognized
action is executed. Keep in mind that the neural network was recognized on my voice samples (11 for each command), so it might not perform good
on you own voice.

"Circle"

"Color"

"Color"

"Triangle"

"Square"

"Clear"

A added function is the "Train 'increase'" which lets the user train the model to pronunciation of the word 'increase' or 'povecaj'.
What it does is makes the shape of the currently displayed shape bigger. The accuracy of the model can't be guaranteed after training
since the weights of the model changed.

"Increase"