• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 100 subscribers
  • Views 1087 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: Voice pre-processing

Takahiro Ogo
Genius 5940 points
Other Parts Discussed in Thread: TIDEP-0077, TIDEP-0083

Hi,

My customer is requesting the demonstration/reference design of Voice pre-processing and network connection on Linux OS.  
Of course, I know TI's reference designs for Voice pre-processing. (TIDEP-0077,0088......)
But these designs is just using DSP only and OS is RTOS. It is not connect to network.

Also, I know "Voice Triggering and Processing With Cloud Connection to IBM Watson® Reference Design"(TIDEP-0083)
But they want 1 chip solution and Bluetooth is required.

Do you have such a demonstration or reference design?
It is OK, ARM + DSP or ARM only. 
I know it is possible if they create, but my customer want to see and run it soon. 
Please advice if possible.

Best Regards,
Taka

  • 0
    TI__Guru**** 393215 points
    The factory team have been notified. They will respond here.
  • 0
    TI__Guru* 84775 points
    Hi Taka-san
    We have an application note that talks about how to hook up a C5517 to AM335x , pointing you to that in case you were not aware
    Sitara Linux ALSA DSP Microphone Array Voice Recognition

    www.ti.com/.../sprac97

    I will ask the authors to provide more guidance on potential single chip solutions, but I do not think we have something up and running in Linux with a unified ARM+DSP processor (K2G does have this as you pointed out - but based on TI RTOS).

    Does your customer have any price target in mind?
    Does the C5517+AM335x solution work?

    Regards
    Mukul
  • 0 in reply to Mukul Bhatnagar
    TI__Genius 13515 points

    Hi Taka-san,

    We have not done much work with ARM-only solutions using a multiple microphone array, but as you have discovered, we have several applications using a DSP for microphone processing.

    Amazon does have a C++ based SDK that includes a sample application a wake word engine from Sensory (triggered by saying "Alexa").  This demo application runs very nicely on the BeagleBone Black and BeagleBone Black Wireless (AM335x running the community-supported Debian operating system + TI Linux kernel).

    More information about the Amazon SDK can be found here: https://developer.amazon.com/alexa-voice-service/sdk.  The source code on github is here: https://github.com/alexa/avs-device-sdk.  The instructions for the Raspberry Pi will largely apply to the BeagleBone, with a few minor differences.  You can find more information on my blog (not affiliated with TI): http://www.erdahl.io/2017/11/alexa-voice-service-on-beaglebone-black.html

    Regards,
    Mike

  • 0 in reply to Mukul Bhatnagar
    Genius 5940 points

    Hi Mukul-san,

    Thank you for your kind support! I understood this situation.

    Mukul Bhatnagar said:
    Does your customer have any price target in mind?
    Does the C5517+AM335x solution work?


    There is no price target yet.
    Currently, they hope to run the voice pre-processing and network/usb on single device(ARM+DSP).
    The following configurations are the best for them.
     - ARM: network/usb on Linux
     - DSP: voice pre-processing on RTOS
     - ARM<=>DSP: SYSLINK, IPC.....
    My customer believe that this configurations can utilize strong points of TI ARM+DSP device.

    If you can do such a demo, please let me know. (I will try to consider this at my end.) 

    Best Regards,
    Takahiro Ogo

  • 0 in reply to Michael Erdahl
    Genius 5940 points

    Hi Mike-san,

    Wow! thank you for your information.
    I wanted to try the AVS on sitara. I will try it! 

    Please let me ask you a question if I can't work it.

    Thanks!
    Taka