Labeling data after recording

After collecting data for a specific class in a separate file, add a class column to the dataset with a constant integer representing that class. Additionally, ensure that the sensor reading columns are properly named.

1. Open the CSV in VS Code, Excel, LibreOffice, or Python.
2. Add a class column.

3. Name sensor reading columns as follows:

   • acc_x
   • acc_y
   • acc_z
   • gyro_x
   • gyro_y
   • gyro_z

4. Label each row according to your gestures.

   For example:

   acc_x,acc_y,acc_z,gyro_x,gyro_y,gyro_z,class
   0.1,0.0,9.8,0.02,0.01,0.00,0
   -0.2,0.5,9.7,0.04,0.05,0.02,0
   -0.3,0.6,9.7,0.05,0.06,0.03,0
   

   Where 0 is the IDLE gesture.

   acc_x,acc_y,acc_z,gyro_x,gyro_y,gyro_z,class
   0.2,0.0,9.8,0.02,0.01,0.08,1
   -0.14,0.5,9.7,0.04,0.05,0.01,1
   -0.3,0.6,9.7,3.05,0.09,0.03,1
   

   Where 1 is the UNKNOWN gesture.

5. Proceed with labeling all 8 gesture classes:

   • Swipe right: 2
   • Swipe left: 3
   • Double knock: 4
   • Double thumb tap: 5
   • Rotation clockwise: 6
   • Rotation counter-clockwise: 7

   Note

   Classes must start from 0. For example, if you are training a model to predict swipe, rotate, tap, encode them as 0, 1, 2. Save the encoding dictionary for future reference.