multi wakeWords support

[alipas777-afk]

Hi again
Is it possible to add a list of wakeWords for training?
It would be great if you added this feature.

[muhammadabidPro]

Hi @alipas777-afk,

Yes, it is definitely possible to support multiple wake words.

The main technical update would be in the BiasWeightedLoss.py file. We need to switch to a multi-class loss function (like Multi-class Focal Loss or Categorical Cross-Entropy) to handle multiple labels simultaneously.

Other parts of the pipeline will only need minor tweaks:

• TTS Data Generation: A small update to generate training samples for each wake word in the list.
• Feature Extraction: Slight adjustments in the feature files to map multiple labels.

Currently, I am working on building an E2E (End-to-End) model to make the system more robust, and I will keep this multi-wake-word feature in mind for the next phases.

Stay tuned!

[alipas777-afk]

Thanks a lot ❤️