Currently I’m using a spectrogram processing block for my audio data, model performs well with it, but I can see that it performs way better if I filter my datasets using a high pass filter (1kHz to remove wind noise). So the question is how I can build a high pass filter in my processing block and use it together with spectrogram? It would be nice to have it like a MFE’s low / high frequency parameters.
You can create your own custom DSP blocks.
We can also host the custom DSP blocks for enterprise customers but you can still use this feature and host them on your servers.
All the existing processing blocks are available this Github repository: https://github.com/edgeimpulse/processing-blocks so feel free to create / adapt your own that exactly matches your needs.
Thank You Louis, I will try it.