• State TI Thinks Resolved
  • Locked Locked
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 771 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.

Original question:

IWR6843ISK-ODS: Low Level Interfacing (uC) with IWR6843

IWR6843ISK-ODS: On-chip Fall Detection Using IWR6843ISK-ODS or WR6843ISK | Improve Human Fall Detection Accuracy

jaydip patel
Prodigy 50 points
Part Number: IWR6843ISK-ODS
Other Parts Discussed in Thread: IWR6843ISK

Hello There ..

We are making product on Smart Health Care for Smart City Project in Vietnam, India and USA , for that we have used the IWR6843ISK-ODS with PC Demo Application. and also i have install CCS on PC and Build own .BIN file and Upload and its working. 

We are using  "ODS_6m_default.cfg"   |    3D_people_count_68xx_demo.bin   |   Fall_detection_Demo UI  |  BoostICBoard  |   IWR6843ISK-ODS

1)  How can we do on-chip fall detection instead of using PC ?

2) How can we improve the accuracy of fall detection ? [Sometimes false-fall detection using PC demo GUI. Is there any alternative sensors that we can use with radar for improving the accuracy ] 

3) How can we implement human-classification on-chip on IWR6843ISK-ODS like IWR6843ISK ? (as IWR6843ISK has human-classification i can see in "sense_and_direct_68xx.cfg" ) 

4) which is the best module for accurate human detection and fall detection ? [ IWR6843ISK-ODS Vs IWR6843ISK ]


Thanks, 
Jaydip 

  • 0
    TI__Genius 14415 points

    Hello Jaydip,

    The PC logic is shown in the visualizer source code, basically looking for a large enough change in the rolling average height of the track over 10 frames. This could likely be implemented in the device. I think somewhere in the MSS_main file would be the best, after the tracker finishes the output of the target values, and before the data is sent out over UART. See the following document for more info on the flow of the 3D people counting demo. 3D_people_counting_demo_implementation_guide.pdf

    The false detections are likely not due to the sensor. The visualizer fall detection logic is fairly simple and meant as an example. Additional filtering on the height and fall conditions, and using the point cloud height instead of the track height could likely improve the offering.

    Currently there is no classification included in the 3D people counting, only the 2D versions (sense and detect). This was implemented before using a k Nearest Neighbor algorithm to classify. This can be implemented in the 3D version, but is not currently included in the demo.

    The ODS board will be more accurate for fall detection. The sensor on this EVM has better elevation angle resolution so it will give a more precise measurement of height. Both boards work, but the ODS typically gives better results.

    Regards,

    Jackson