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.

cc2350 video recognition

Other Parts Discussed in Thread: CC2530, CC2520

Dear all,

 

I want to do a project based on video recognition using c2350, The idea is that instead of inputting some digital or analog data to the cc2350, input a video recorded by a small digital camera, and then using some sort of software that is imbedded on the chip that does video recognition, certain actions might be taken depending on what the camera sees. The 2nd part is simple, but the first part is a bit complicated.

 

I have been using this kit for more than 6 months now and I have never encountered something like this, not even on the forums here.

So my question could be as following, If I by any means was able to input a video to my device using the I/O pins, will the kit be able to handle such a thing??

 

This is very complicated I know but if anyone could help, I would be grateful.

 

Thanks

 

 

 

 

  • Hi Omar,

    First of all, I based my whole answer assuming that the device in your post is CC2530,

    otherwise, the answer isn't relevant.

     

    Now, I might be not able to help you, however one of the engineers in my firm

    doing some experiments with voice over Zigbee on top of Zstack running

    on cc2530 SoC. Which means that some audio stream (compressed, or not)

    should be received from some CODEC by the means of SPI and stored in cc2530

    RAM for further processing.

    At this stage this whole thing seems like a huge challenge, since cc2530 SPI<->I2S

    interface limiting us to a very low bit rates (8Khz sample frequency is the highest option).

    Moreover I'm not sure if cc2530 capable of receiving continuous stream of data and

    processing it, as the 32MHz system clock is relatively slow for such kind of tasks. In addition,

    a device based on cc2530 with running Zstack on it is very limited on RAM (~4Kbytes of free

    RAM in the best case).

     

    In the bottom line, I have a big interest in scenario where the cc2530 SoC handles an audio/video

    stream + doing some processing and sending the processed data to ZigBee Concentrator.

    My guess, it won't work! However, maybe some sort of efficient algorithm and a camera with

    integrated compression device can do the trick.

     

    Br,

    Igor.

  • Dear Igor,

     

    Thanks for your reply. I think it says it all. I was kind of thinking the same also but an idea came to my mind a few hours ago. The idea is simply as follows, the camera is attached to any end device, and this end device does no processing at all, only gets the streaming and forward it to the sink where the sink forwards it to a computer, where all the processing done.

     

    But this also seems not to work.

     

    I will try to think about another thing to do, anyways thank you very much for your time.

     

    Omar

  • Dear Omar,

    In my opinion, the best configuration for such device (camera<->cc2530) is

    when the application running on ZigBee end device, however, probably you

    are right, even in this case it won't work.

     

    Actually I have another idea.

    what if you'll be using TI's MSP processor along with cc25xx ZigBee device

    and the camera connected to the MSP. The whole image processing stuff

    will be carried by the processor and the processed data sent to the ZigBee

    network through the cc2520 (for instance) device.

    What you think?

     

    Br,

    Igor

     

  • I would be highly skeptical of achieving any video recognition capabilities using either an CC253x or an MSP430 part.  Mostly due to RAM constraints but also due to processing horse power.  Not that id couldn't be done but it certainly couldn't be done real time.  The delay in actually achieving such an algorithm on either of these processors would be significant (several seconds to minutes per frame) to say the least.  If that is acceptable, then the next biggest issue is in storing the data as these processors have very limited amounts of RAM and neither have a direct access mechanism to a standard asynchronous or synchronous SRAM and interfacing to a DRAM would be a piece of work as well.

    I think Igor has the right idea here which is to use another DSP like a 55xx or 6xxx part or even better to us a DM6x part with a video codec built in.  Then utilize a radio front end to achieve management of the event the video processing would eventually generate.  Such a platform has the ability to achieve real time operation.

    Jim Noxon

  • After searching and discussing the idea with my supervisor, We finally have found the best solution. We are getting Davinci zigbee kit that has a DSP and built-in camera that takes the inputs. We think this kit can be used to do this project. So I want to thank you all for your attention and help.

     

    Best regards

     

    Omar

  • Hello Omar,

    Another point to consider is the low throughput of a ZigBee Network around 20KBps, mostly a deterrent in streaming audio and video on a ZigBee network which require continuous streaming at high data rates.

    Regards,

    Suyash Jain

  • Hi Omar,

     

    Can you tell me please the name of this kit, or insert a link

    to this kit on TI website.

     

    Thanks in advance,

    Igor

  • Its called EDK6446 TI DaVinci based Development Kit


    here is a link on TI website: http://focus.ti.com/dsp/docs/thirdparty/catalog/devtoolsproductfolder.tsp?actionPerformed=productFolder&productId=4237


    Its amazing. Actually my supervisor has one already, but he has never used it. I will be using it very soon

     

    BR

    Omar