is there any particular reason why I cannot get the system to recognise more than 3 individual words?
**Project ID:308092
**I am trying to get my Nicla Voice to recognise 3 words with just my voice:
1 - ok bot
2 - chiama (call)
3 - sali (up)
4 - scendi (down)
I got it to recognise words 1,2,3 but if I add even a few samples of word 4a it is recognised but word 3 is no longer recognised.
I have tried using both optimisations and without.
If I delete word 4 the others start working again.
If I create a dataset for a different 4th word this one works and again the 3rd is no longer recognised.
Is this a problem with the quality of the dataset or are there system limitations?
It looks like you have an unbalanced dataset with, for example, 10 sec of “sali” and 101 sec of “scendi.” The model needs more data to learn the differences between these words. I recommend getting a lot more (try to get at least 500 samples, 1 sec each, of each of your classes). I also recommend balancing your dataset so that you have a nearly equal number of samples in each class (e.g. 500 samples for each class).