Has anyone else seen this error? Attached the terminal output.
NemoDockerErrorSampleRate (23.5 KB)
Hi @Danieljyu2n, the Quartznet ASR model uses 16KHz sampling rate, and it tries to open the microphone at this rate. I haven’t yet implemented resampling so that it can take a 44.1KHz mic and downsample it to the needed 16KHz. It appears your microphone only supports 44.1KHz.
What microphone are you using? I would like to get one that only supports 44.1KHz mode and not 16KHz so that I can test this. In the interim, you may want to try a mic that supports 16KHz mono.
Hi Dusty, I have been researching this and trying to work through this because I believe the mic can handle 16KHz. Before I load the container I can run audacity and record and playback 16KHz files via the USB audio in/out adapter. In addition while in the container, I can play your dusty.wav file using aplay if and only if I use the -D sysdefault:CARD=2 option. If I don’t I get an error. It is my guess that the container does not have the correct ALSA settings. I tried to add an asound.conf to etc and then force an ALSA reload, but that didn’t help. Let me know if I am missing something obvious as containers are new to me.
Thanks, Daniel
PS. This is one of those 3D USB adapters you get on cheese-bay for $3.00
The ALSA settings the mic is opened with are found here in the code:
You can start the container in --dev
mode and edit this file in your jetson-voice repo in the host. It needs to be 16KHz mono for the ASR to work. I think maybe you could try changing the int16 to float32.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.