Data Augmentation of 1D data

TensorFlow
, , , ,
1

I am doing Hand Gesture Recognition using sensor readings from 10 flex sensors. How can I augment the data, to get good accuracy? Because currently, I am getting only 20% accuracy.

|S.No. S1 S2 S3 S4| S5 |S6 |S7 |S8 S9 |S10 |Cls|
|1| 577.87 2989.01 1777.78 7003.1 4566.8 1965.87 67200.01 -9002.2 2142.86 935.72 |1|
|2| 577.87 2989.01 1934.17 7151.33 4656.72 1965.87 3081.08 -9002.2 2126.58 922.1 |1|
|3| 616.65 2989.01 1902.6 7241.38 4634.15 1965.87 3007.34 -9002.2 2126.58 908.52 |1|

My dataset consists of 10 columns and approximately 12 samples for each hand gesture. Kindly help

3

Hi @Vevinya_A ,

Here are some techniques that you can try:

  1. Random scaling
  2. Random translation
  3. Random rotation
  4. Adding noise
  5. Mirroring
  6. Time shifting

You can experiment with different combinations of data augmentation techniques to find the ones that work best for your dataset. Collecting more data if possible, as a larger dataset can also help improve the accuracy of your model.

Please let me know if it is answer your question.

Thanks.

5

Yes, it did. Thank you mam.