Speech to text Solution with i.MX93 SOM

[DeepThinker]

Hi, I am trying to bring up Speech to Text solution with i.MX93 processor.

I have found one in the github, which is compatible to i.MX93. Location : https://github.com/nxp-imx/nxp-afe?tab=readme-ov-file.

I was able to clone and build and got the required files and executables. The steps to execute is also written in this link: https://github.com/nxp-imx/nxp-afe/blob/MM_04.07.02_2210_L5.15.y/TODO.md 

I am little confuse on the how to execute program section. it says 

1. install aloop module to support afe insmod snd-aloop.ko

2. handle asound.conf You'd better make a backup of /etc/asound.conf in rootfs, and then copy /unit_tests/nxp-afe/asound.conf to /etc/ to replace the asound.conf file there. There is a little different on imx8mp board because the config file is asound.conf_imx8mp. Apart from copying it to /etc/, the name should also be changed like: cp /unit_tests/nxp-afe/asound.conf_imx8mp /etc/asound.conf

3. cd to /unit_tests/nxp-afe/ execute afe refering to the following command: ./afe libdummy ./afe libdspc ./afe libfraunhofer ./afe libvoiceseekerlight                             where exactly do we need to install aloop? second point says for imx8m, so for imx93, do we just need to replace with imx93?

first of all. am I in the correct repository which can help me bringing up speech to text with imx93 processor? can anyone help me how to run the executable on i.MX93?

#speechtoaudio #iMX93

#NXP @Tia_Lan  Harvey021

@L30 @JorgeCas 

[L30]

@DeepThinkerHi !

I think that in order to use the snd-aloop module you need to enable it in your kernel config and recompile it

take a look here https://github.com/instantinfrastructure/linux-yocto-3.10/blob/a3fca6aa42adc1dcbad019ac91515e98a43dc...

Regarding the speech to text , I've never use it so I'm really sorry but i think I cannot help you ...but that's an interesting topic so i will follow this post

[JorgeCas]

Hello, I hope you are doing well.

Yes, you are in the correct repository, maybe you should change the branch according to your Linux kernel.

For the three steps you need to initialize the module and replace the files to indicated directories.

Best regards.

[DeepThinker]

Hi Jorge, Thanks for your response and clearing that I am in the right repository.

I have two doubts,

1. I am not able to install the aloop module, it gives this error.

2. Can you explain where, when and who exactly is converting the Speech to text in the block diagram?

Thanks in advance.

[JorgeCas]

Hello,

Maybe you could try with:

sudo modprobe snd-aloop

The Audio Front End repository does not detail in depth where is converting the Speech to text. But, according to the description, once filtered microphone samples are obtained are stored, so the applications can have access to it and work with that data.

Best regards.