• State TI Thinks Resolved
  • Locked Locked
  • Replies 13 replies
  • Answers 5 answers
  • Subscribers 99 subscribers
  • Views 5820 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Linux/TMS320C5535: Customizing the Trigger Phrase

Osmar Resendiz
Prodigy 250 points

Part Number: TMS320C5535

Tool/software: Linux

Hello, everyone

I've been trying to customize the trigger phrase on the C5535 DSP card as the Reference Guide says but with no luck whatsoever. I'm completely new to Linux, so that's a problem. Regardless, I think I have followed the instructions on the guide to the letter but a few problems showed up. (Also I think it's important to say that I am using Linux on a bootable USB. I don't know if that has something to do with anything, just wanted to clarify that.)

On the subsection 5.3.3 the number 3 says "Execute the following command: $make LinuxDebugGnu", by this I assume I only have to write "make LinuxDebugGnu" for the symbol $ is already there. So I did that and I got the next message on the terminal:

I think it didn't worked because the Dist directory is not created as the next sentence says. No "bin" folder was created and the "lib" folder is empty as shown below.

Any suggestions? Or does someone know another (preferably easier or not involving Linux) to customize or change the trigger phrase on the DSP? I heard it might be possible in MatLab. Any comments?

I'd appreciate any advice.

- Osmar

  • 0
    TI__Guru 65405 points
    Hi Osmar,

    I've forwarded this to the c55x software experts. Their feedback should be posted here.

    BR
    Tsvetolin Shulev
  • 0
    TI__Mastermind 30365 points
    Osmar,
    Please see this previous thread on a similar issue e2e.ti.com/.../523856

    The user was able to eventually get past the issues after switching to a different version of Linux. Not sure why though.
    I had also attached the Dist folder to unblock the user. Perhaps this would help you.

    Lali
  • 0 in reply to Lalindra Jayatilleke
    Prodigy 250 points
    Lalindra,

    I can't open the link, it says the access is denied maybe because of the thread was deleted. I would really like to have a more direct way o communication, Lalindra. Would it be possible for you to give me an email address. Or for you to contact me on a personal email address? I have some rather deep doubts about the DSP I'd like to discuss with you.

    - Osmar
  • 0 in reply to Osmar Resendiz
    TI__Mastermind 30365 points

    Osmar,

    E2E is the only route of communication at the moment. My apologies. Please let us know what the doubts are and we can try to assist here.

    To help you along, I have attached the DIST folder for the build. Is there another version of Linux you could try to build on?

    Lali

    /cfs-file/__key/communityserver-discussions-components-files/791/6567.TIesr_5F00_model_5F00_build.zip

  • 0 in reply to Lalindra Jayatilleke
    Prodigy 250 points

    Lalindra,

    Thank you so much for your attention. I suppose I could try to download a previous version of the software. Which on do you recommend? Also, what if I want to change the "actions" of the DSP, for example: Now, when the trigger phrase is detected it appears on the screen. How do I make it to turn on a led, or change the phrse displayed on the screen, or send a TTL pulse for example. Do I need to change the C code? Or is it no even possible? That is actually what I bought the DSP for, it's for a final project at school, so it would really be a bummer if it was not possible. A little help in that matter would be immensely appreciated.

    Also, with that Dist folder you just provided me with, do I not have to make the whole "make LinuxDebugGnu" thing now? Can I just paste the folder and proceed as the Reference Guide says?

    - Osmar

  • 0 in reply to Osmar Resendiz
    TI__Mastermind 30365 points

    Osmar,

    If you look at recognizer.c, there is a function Void swi_RecognizerFxn(Void)


    You could put your action code where you print to the OLED. So essentially tack onto the action of printing to the OLED to call a TTL routine. Keep in mind that BIOS is managing a few tasks in the background, so you would have to ensure correct priorities are set if you are dealing with a peripheral and interrupts.

                            /* Output answer to OLED LCD */
                        print_string((char *)word, i2cHandle);
                        print_string(" ", i2cHandl

    You could try with Ubuntu 12.04. And, yes on the DIST folder. YOu should be able to drop that in and continue. Hope this helps.

    Lali

  • 0 in reply to Lalindra Jayatilleke
    Prodigy 250 points

    Lalindra,

    That's fantastic to know! But, in which particular pin would I be able to get that TTL pulse? Or what terminal do I use to output voltage, whether it is TTL or just DC to light a led? Also, the phrase displaying on the OLED display is whatever I change it to on Linux, right?

    - Osmar

  • 0 in reply to Osmar Resendiz
    Prodigy 250 points
    Lalindra,

    I was just thinking about this, is it possible to have a "two step" recognition? For example, to be able to activate the system by saying "t i voice trigger" and THEN saying another instruction like "light on" or something like that to turn on a led like I asked before. The first phrase will be something like an "activation key" and the second one would be the instruction. Any idea on how to do this?

    -Osmar
  • 0 in reply to Osmar Resendiz
    TI__Mastermind 30365 points
    Osmar,

    It sounds to me like you want wake-word detection followed by processing of an actual command.
    The wake-word part is something you have already figured out. Unfortunately, I'm unaware of a multistep local keyword recognition implementation.

    Something else you could try is cloud based voice transcription like what Google or IBM provides. You can use the wake-word to trigger the listening and then record any subsequent command. This audio clip can then be sent to the cloud for transcription and you could conduct some action based on the returned text. You would need some Internet connectivity in your system for this.

    Lali
  • 0 in reply to Lalindra Jayatilleke
    Prodigy 250 points
    Lalindra,

    Well... the think is I was really hoping to do that on the DSP, not needing to rely on a cloud service... Is thera a way for me to combine both techonoogies? Implementing the DSP with a cloud service, or would that just be redundant?

    - Osmar
  • 0 in reply to Osmar Resendiz
    Prodigy 250 points
    Sorry, I meant thing* on the first sentence.

    - Osmar
  • 0 in reply to Osmar Resendiz
    TI__Mastermind 30365 points
    Osmar,
    You can certainly integrate the DSP with a cloud service, but you would need another device such as a CC3220 wifi module to handle the connectivity. As I had mentioned earlier, we don't have an example showing a multistep voice recognition system.

    You could perhaps try experimenting in the code to see if a multistep recognition system can be built.

    Lali
  • 0 in reply to Lalindra Jayatilleke
    Prodigy 250 points

    Lalindra,

    Thank you for your attention, I I'll just star working on the code then. I'll post here for any other doubt.

    - Osmar