mmWave radar solution for human Fall Detection

Nikos Papastefanou

Other Parts Discussed in Thread: IWR6843AOP, MMWAVEICBOOST, IWR6843

A company asked from us to develop for them a Fall detection system that will be non-wearable, in order to be manufactured and installed for remote patient monitoring. .

So we thought to implement a solution based on mmWave radar.

But , unfortunately its a very new technology , I am not familiar with radar systems and i would like to know if this can be achieved using TI mmWave radar systems.

Thank you in advance

over 2 years ago

0 Nathan Block over 2 years ago

TI__Mastermind 37805 points

Hi Nikos,

This is absolutely something that can be achieved with radar technology.

My suggestion would be to order an IWR6843ISK EVM, and run the 3D People Counting Code on it. The newest people counting code has a height estimation feature for each detected person that can be used as a basis for fall-detection. There are also many other ways to do fall detection as you may find in a state-of-the-art literature analysis.

Best,

Nate

0 Nikos Papastefanou over 2 years ago in reply to Nathan Block

Prodigy 30 points

Thank you for your suggestion Nate.

The problem is that my client demands short delivery dates. For example they asked a biopatch that should be delivered within 3 months , so i had to use a more expensive AFE that was providing sdk and calculation algorithms instead of one of your trusted AFE’s and seems that i will deliver on time.

However, in this case my client demands delivery on next 45 days, otherwise he will rebrand an existing product from China. So, i have no time for R&D, but only to put things together and deliver a product with BLE and WiFi connection. So realistically, I should use kind of a ready solution. I don’t know if ts allowed to mention other brands here on this forum, but there is a hardware solution with ready firmware with specific outputs, ready to communicate with a MCU. However, if I could have support on firmware and pcb review I would prefer TI.

0 Nikos Papastefanou over 2 years ago in reply to Nathan Block

Prodigy 30 points

Dear Nathan,

I saw a video demonstrating fall detection using a demo for IWR6843AOP. Its also mentioned a demo called " Elderly fall detection " with redirection on following link

dev.ti.com/.../node

On that link, i didn’t find that demo, but i found a demo called "Vital Signs With People Tracking", which is also on plans to be integrated as function.

As i said, if i agree to provide to my customer a product, this will have to be delivered on next 45 days,

so will especially my firmware engineer regard the related algorithms? Or we have to develop from scratch?

0 Nathan Block over 2 years ago in reply to Nikos Papastefanou

TI__Mastermind 37805 points

Hi Nikos,

The demo you're referencing can be found here

https://dev.ti.com/tirex/explore/node?a=1AslXXD__1.00.00.26&a=VLyFKFf__4.10.0&a=VLyFKFf__4.10.1&a=VLyFKFf__4.6.0&a=VLyFKFf__4.7.0&a=VLyFKFf__4.8.0&a=VLyFKFf__4.9.0&node=A__AP4ttSY7mvWfdzU56Gb8ZQ__com.ti.mmwave_industrial_toolbox__VLyFKFf__4.10.1

Best,

Nate

0 Nikos Papastefanou over 2 years ago in reply to Nathan Block

Prodigy 30 points

If we get the ev kit and the MMWAVEICBOOST could we have access to the Fall Detection Algorithm source code in order to run this on a microcontroller or microprocessor if more process power needed?

+1 Nathan Block over 2 years ago in reply to Nikos Papastefanou

TI__Mastermind 37805 points

Hi Nikos,

As the page I linked above states : the demo runs on the 3D_people_count_68xx_demo binary, and the fall-detection algorithm runs on the python visualizer. The entire algorithm src is present.

Additionally, our most updated software offering for 3D People Counting, found in the radar toolbox version 1.0, contains a height detection feature for each track that runs on the IWR6843 device. This could also be used to develop a fall-detection algorithm.

Best,

Nate

0 Nikos Papastefanou over 2 years ago in reply to Nathan Block

Prodigy 30 points

Dear Nathan,
do you know if the algorithm can run on sensor's mcu so that it outputs the state instead of raw data?
Thanks

0 Nathan Block over 2 years ago in reply to Nikos Papastefanou

TI__Mastermind 37805 points

Hi Nikos,

Yes the algorithm could be run on the MCU instead to output state instead of just raw data.

Best,

Nate

Sensors

Sensors forum

mmWave radar solution for human Fall Detection