Skip to main content
J
JamesE
Visitor II
March 8, 2022
Solved

Can ASM330LHH eliminate most or all false detections?

  • March 8, 2022
  • 1 reply
  • 1073 views

Hello experts,

I'm trying to determine if the ASM330LHH IMU can eliminate false positives by itself or if the MCU will need to address them. I've searched the forum for this topic and didn't find anything specifically related to this device. Apologies if I missed any.

I did read through the ASM330LHH: Automotive inertial module digital 3D accelerometer and digital 3D gyroscope app note (AN5296). It is a helpful document, but I didn't find much about how to prevent or handle false detections.

Page 30 includes the following sentence that indicates that false detections can be avoided by using a longer duration.

The FF_DUR[5:0] field of the FREE_FALL / WAKE_UP_DUR registers is configured like this to ignore events that are shorter than 6/ODR_XL = 6/412 Hz ~= 15 msec in order to avoid false detections.

In my application, I'm more concerned about detecting certain impacts rather than free fall scenarios. Thus, I would assume setting an accelerometer magnitude threshold and a duration threshold (similar to what's illustrated in Figure 13) would be enough to avoid false detections.

Is my understanding correct? Thanks for your help in advance.

James

    This topic has been closed for replies.
    Best answer by Eleon BORLINI

    Hi @JamesE​ ,

    yes, your understanding is in principle correct.

    If you want to discriminate a specific signal (with a specific amplitude and a specific duration), that is different from all the other signals and cannot be confused with them, you can use the free fall, or the wake up features (asm330lhh_free_fall.c / asm330lhh_wake_up.c).

    If otherwise the amplitude and the duration are not enough to discriminate, you might switch to devices with embedded Machine Learning Core (MLC), such as the ISM330DHCX: MLC allows the user to select among more complex features than duration and threshold, since you can use mean, variance, RMS, etc to build a decision tree that can classify your signal.

    If my reply answered your question, please click on Select as Best at the bottom of this post. This will help other users with the same issue to find the answer faster.

    -Eleon

    1 reply

    E
    Eleon BORLINIAnswer
    ST Employee
    March 10, 2022

    Hi @JamesE​ ,

    yes, your understanding is in principle correct.

    If you want to discriminate a specific signal (with a specific amplitude and a specific duration), that is different from all the other signals and cannot be confused with them, you can use the free fall, or the wake up features (asm330lhh_free_fall.c / asm330lhh_wake_up.c).

    If otherwise the amplitude and the duration are not enough to discriminate, you might switch to devices with embedded Machine Learning Core (MLC), such as the ISM330DHCX: MLC allows the user to select among more complex features than duration and threshold, since you can use mean, variance, RMS, etc to build a decision tree that can classify your signal.

    If my reply answered your question, please click on Select as Best at the bottom of this post. This will help other users with the same issue to find the answer faster.

    -Eleon

    J
    JamesEAuthor
    Visitor II
    March 12, 2022

    Hi @Eleon BORLINI​,

    Thanks for your detailed reply. Your feedback has answered my initial questions, and hopefully our future testing will clearly demonstrate that amplitude/duration is adequate or not enough for our application.

    Have a great weekend.

    James

    Terms & ConditionsAccessibility statement