Supervised Expert System for Wearable MEMS Accelerometer-Based Fall Detector

Pseudocode 1

Feature extraction pseudocode.

Let 𝐴₀ initial acceleration value measured after the calibration phase

Let 𝛼𝐴₀ initial axis angle of the sensor measured after the calibration phase

Δ𝐴  →  shock dynamics due to the impact calculated in the first part of the sliding window

𝐴  →  Static averaged acceleration in the third part of the sliding window

If compensation device misplacement procedure is enabled

  𝛼𝐴  is calculated → axis angle of the sensor in the third part of the sliding window

  CPO=sin𝛼𝐴−𝛼𝐴₀ → Changing Position Offset

else

  CPO=𝐴−𝐴₀ → Changing Position Offset

End if

Let Δ𝐴_𝑥, Δ𝐴_𝑦, Δ𝐴_𝑧, CPO_𝑥, CPO_𝑦, CPO_𝑧, Δ𝐴 and CPO calculated for each axis

[𝑋,𝑌,𝑍]=[Δ𝐴_𝑥⋅CPO_𝑥,Δ𝐴_𝑦⋅CPO_𝑦,Δ𝐴_𝑧⋅CPO_𝑧] → feature vector input for  SVM