• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 53 subscribers
  • Views 388 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.

CCS/IWR1642BOOST: Identify fall detection in 16xx using 2D point cloud data

Vivek Marimuthu
Prodigy 150 points
Part Number: IWR1642BOOST
Other Parts Discussed in Thread: IWR1642, IWR6843

Tool/software: Code Composer Studio

Fall Detection:

Could you please tell me the feasibility of understanding fall detection using IWR1642.

  1. Currently I am using IWR1642 device for tracking people.
  2. With the help of Raspberrypi , oob_parser and 2D People counting, I am getting 2D people count data of a person.

Scenario 1:

At T time, person will be standing in front of sensor, using 2D people cloud point data, I have to calculate dimension of the that person.

From forum, I got to know that height of the person will be calculated by "Sensor's height - Measured Length"

Query #1:

How to calculate Length / dimension of the person using 2D point cloud data? 

Is there any oob_parser available for “Level Sensing” (To identify the distance from sensor)?

Scenario #2:

At T + n time, person is falling down

Query #2:

When a person fall down, how do I identify fall detection using cloud points?

Please clarify.

Regards,

Vivek

 

  • 0
    TI__Mastermind 28055 points

    Hi Vivek,

    We have not tested a fall detection demo using 2 dimensional data. From the experiment we have on TIREX, you can see that used a 3D Point cloud. 

    My recommendation would be either:

    1. Save point cloud data for a person standing, walking, sitting, and falling. You will not be able to use the X Y coordinates to help you, but analyzing the doppler component of the point cloud may make it easy to find when a person is falling.
    2. Get an IWR6843 ISK or ODS evm to enable 3D data.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 150 points

    Hi Justin,

    Thanks for your response.

    We already having IWR6843 device. As suggested, 3D point cloud data can be retrieved using out of box demo. That we understood.

    But our concern here is, with the point (2D/3D) cloud data how will I arrive height / dimension of the person?

    We saw some experiments related to Fall detection in TI forum (Detecting Human Falls and Stance with IWR6843ODS - http://dev.ti.com/tirex/explore/node?node=AFmsC4Pjv3N8OtdrcDLsEg__VLyFKFf__LATEST&search=mmWave%20industrial%20toolbox)

    Can we have any demo application with the source code?

    Please share

    Regards,

    Vivek

  • 0 in reply to Vivek Marimuthu
    TI__Mastermind 28055 points

    Hi Vivek,

    The tracker returns a TLV called track index. This is a mapping between tracks and points -> tracker index is a array of uint8, index corresponds to a point. So if tracker index[4] is 2, then pointCloud[4] is associated with track 2. Please note that point cloud output is always frame N, and tracker output is always frame N-1, so you need to save the point cloud for one frame. 

    With the above information, you can get all of the points associated with the track. For height, take the highest point for the track, and average over 10 frames. A fall is detected when there is a large change in height. You could use similar techniques to try to estimate length, width, height. This technique works well for fall detection because we don't care about the true height of the person, just the relative change in height. 

    Regards,

    Justin