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.

IWR1443 or IWR1642 which EVMs can do people fall detection in the bathroom?

Other Parts Discussed in Thread: IWR1443, IWR1642, DCA1000EVM, IWR1642BOOST, IWR1443BOOST

Our company want to make a bathroom fall detection product for smart homes as soon as possible. We found that the "Very Fine Motion Detection vs PIR" tutorial from Texas Instruments says that the following behaviors are identified using the IWR1443 and IWR1642 respectively.

  1. Major motion: Person walking, Arm movement, etc.
  2. Fine motion: Typing
  3. Very fine motion: Breathing

http://dev.ti.com/tirex/#/?link=Software%2FmmWave%20Sensors%2FIndustrial%20Toolbox%2FExperiments%2FVery%20Fine%20Motion%20Detection%20vs%20PIR

For this tutorial I have the following questions:

1. Which mmWave radar IWR1443 or IWR1642 should I use for bathroom falls detection?

2.  The tutorial “Very Fine Motion Detection vs. PIR” contains information about people's behaviors such as walking, standing, arm movements, typing, and breathing. Does thie demo can do the people fall detection?  If not, can fall detection be done with simple modifications? Do you have any suggestion for this modification?

3. For our fall detection requirements scenario, should we purchase that development board to develop IWR1642-DATA-CAPTURE-BNDL or IWR1642BOOST?  What's the difference between this DATA-CAPTURE-BNDL and BOOST? Does it mean that if we do not purchase the DCA1000EVM there is no way to modify the SDK and algorithms?

4. I also found out that there is a demo for People Counting User's Guide (http://dev.ti.com/tirex/#/?link=Software%2FmmWave%20Sensors%2FIndustrial%20Toolbox%2FLabs%2FPeople%20Counting%20Demo)

So what do we do to do fall detection based on the demo of this people counting? Or is it based on the demo of Very Fine Motion Detection? That is faster and we hope to introduce prototype products as soon as possible.

  • Hi Weisen,
    The different between IWR1642 and IWR1443 is that 1642 contains DSP instead of Hardware Accelerator in 1443. IWR1642 is more flexible on algorithm development.
    In fall detection application. It's better have elevation information to judge if people is fall or not. (IWR1443BOOST have the elevation antenna array, IWR1642BOOST only have azimuth array)
    If you need to develop your own algorithm, it's better capture ADC RAW data using IWR1642BOOST/IWR1443BOOST + DCA1000. Then analyse the data and run your own algorithm on PC. After the algorithm is done, you can translate it to C code run on device.
    The signal chain run on the device are similar. You can start with SDK out-of-box demo and capture some 3D point cloud data in fall detection application.
    If you have any further questions, please let me know. Thanks.
    Regards,
    Wesley
  • Thank you for your quick respond.

    For this fall detection prototype, we hope have a quick solution. We don’t want develop our own algorithm currently. For this request, according to your suggestion we should buy the IWR1443BOOST. In my understanding of your suggestion, we should follow the below steps.

    1. The IWR1443 EVM is running the mmWave SDK 1.1 Out-of-Box Demo.
    2. The windows PC is ruing the Vital Signs Demo.
    3. From the mmWave SDK 1.1, we can get the elevation of each person. And then we will use the person elevation to decide this person stand or fall.
    4. Furthermore, if we want improve the accuracy. Because we can get the 3D point cloud data of each person form mmWave SDK 1.1. And then we design algorithm to identify the fall based on each person’s 3D point could.

    For this steps, I have follow question:
    1. Does the mmWave SDK 1.1 Out-of-Box Demo can get elevation of each person?

    2. Does the mmWave SDK 1.1 Out-of-Box Demo can get 3D point could of each person? Do we need to further segment of each human?

    3. We don’t need buy the DCA1000. Because we use the mmWave SDK 1.1. we don’t do any modify of SDK. We don’t use the ADC RAW data. Is this understanding correct? 

    4. The function of DCA1000 is get the raw data? If we want design the algorithm form from scratch in the future, we need buy the DCA1000. For current condition, my question is for this fall detection prototype do we need buy the DCA1000?
  • Hi Weisen,

    The answers are listed below:

    1. Does the mmWave SDK 1.1 Out-of-Box Demo can get elevation of each person?

    Yes, Out-of-Box Demo can get elevation information of each object.

    2. Does the mmWave SDK 1.1 Out-of-Box Demo can get 3D point could of each person? Do we need to further segment of each human?

    Yes, Out-of-Box Demo can get 3D point could for each object. Here's some demo show video link you can refer.

    https://training.ti.com/robotics-sense-and-avoid-demonstration-using-ti-mmwave-sensors?cu=1135109

    https://training.ti.com/point-cloud-visualization-robotics-using-tis-mmwave-sensor

    3. We don’t need buy the DCA1000. Because we use the mmWave SDK 1.1. we don’t do any modify of SDK. We don’t use the ADC RAW data. Is this understanding correct?

    Yes, if you don't want to get the ADC RAW DATA, you didn't need to but DCA1000

    4. The function of DCA1000 is get the raw data? If we want design the algorithm form from scratch in the future, we need buy the DCA1000. For current condition, my question is for this fall detection prototype do we need buy the DCA1000?

    Yes, if you want to base on point cloud, and do some high level algorithm to identify if people is fall or not. You don't need DCA1000. If you want to improve the point cloud, you may need the DCA1000 to capture the ADC RAW data.

    Regards,

    Wesley

  • Thanks You very much. It definitely helped. 

    I have one more question:

    Does the IWR1443BOOST EVM can run People Counting demo?
    dev.ti.com/.../
  • Hi Weisen,

    People counting demo can only run on 1642 platform.
    Thanks.