In order to acquaint myself with pocketsphinx, I've gotten it to run pretty well on both Windows (Visual Studio 2017) and a Raspberry Pi using the default acoustic model en-us.
Now I would like to get it to run on an a 32-bit microcontroller with limited memory (2 MB Flash, 2.5 MB RAM) compared to the Raspberry Pi. I only need to recognize about a dozen words, in two or three word groups for controlling audio playback (like "Volume Up", "Volume Down" etc.). It has both a USB host interface (for dongles) and SD card interface, but unfortunately the processor and file system don't support memory-mapped I/O.
Is there a way to significantly reduce the size of the existing acoustic model files for a small vocabulary without having to create new ones from scratch? I went through the steps in the tutorial for "Adapting the default acoustic model" and thought it would generate new, smaller files but it didn't.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
In order to acquaint myself with pocketsphinx, I've gotten it to run pretty well on both Windows (Visual Studio 2017) and a Raspberry Pi using the default acoustic model en-us.
Now I would like to get it to run on an a 32-bit microcontroller with limited memory (2 MB Flash, 2.5 MB RAM) compared to the Raspberry Pi. I only need to recognize about a dozen words, in two or three word groups for controlling audio playback (like "Volume Up", "Volume Down" etc.). It has both a USB host interface (for dongles) and SD card interface, but unfortunately the processor and file system don't support memory-mapped I/O.
Is there a way to significantly reduce the size of the existing acoustic model files for a small vocabulary without having to create new ones from scratch? I went through the steps in the tutorial for "Adapting the default acoustic model" and thought it would generate new, smaller files but it didn't.
You'd better check something like
https://github.com/ARM-software/ML-KWS-for-MCU