In-Vehicle Baby Alarm App; Figure 24. Human Activity Recognition App - ST STEVAL-MKSBOX1V1 User Manual

Four features are used (mean, variance, peak-to-peak, zero-crossing) for MLC code generation.
The algorithm runs at 26 Hz, with a window of 75 samples.
The Human Activity Recognition app works with the following ST high accuracy MEMS acceleration sensor and
operating parameter settings:
• APP DATA INPUT: Accelerometer values
• SENSOR USED:
• SENSOR SETTINGS:
– Low power mode
– Output data rate: 52 Hz
– Low pass filter: 700 Hz
– Full scale: 2 g
• APP DATA OUTPUT :
– Activity recognized icon
3.1.6

In-Vehicle Baby Alarm app

The In-Vehicle Baby Alarm app combines the state of baby crying (see
movement detector.
The sensors used are the
gyroscope, as well as the embedded Machine Learning Core.
For the MLC code generation, the following features calculated from accelerometer and gyroscope values have
been used: MEAN-acc, VAR-acc, PeakToPeak-acc, MAX-acc, MEAN-gyro, VAR-gyro, PeakToPeak-gyro, MAX-
gyro, MIN-gyro, ENERGY-gyro.
The app shows:
• whether the adult is in-vehicle or not
• whether the baby is crying or not
• the alarm icon if there is no adult in vehicle and the baby is crying
UM2580 - Rev 5
Figure 24. Human Activity Recognition app
LSM6DSOX (high bandwidth acceleration sensor)
MP23ABS1 analog MEMS microphone and the
Mode 1 example apps
Section 3.1.1 ) and the vehicle
LSM6DSOX MEMS accelerometer,
UM2580
page 21/41