How to implement speech recognition on jetson nano


we want to implement a speaker recognition pipeline on jetson nano using the expert API and a raspberry pi connected to the AIY kit. Are there ways we could do this easily?



Please check the below voice demo container first.


The voice container is for Xavier (needs Volta GPU), however you can try the NeMo toolkit on GitHub:

There is also a forum thread about building it for Jetson/ARM:

I have been hoping that a new working jetson voice container would run under the current jetpack 4.51 or the proposed last jetpack 4.6. Alas this will never happen due to the lack of Nvidia support for the Jetson developer product. I have found a work around. The third party project picovoice is running on my Jetson NX. It supports a wakeup command and additional commands after wakeup. If you are building a robot using the Jetson this will work. If you have problems getting it to run please post them here and I will try to help. The few glitches I found were promptly answered by the picovoice staff and are posted in the picovoice forum. Actual customer service…go figure!