Colocalized Sensing and Intelligent Computing in Micro-Sensors

. 2020 Nov 6;20(21):6346. doi: 10.3390/s20216346

Algorithm 2 Postprocessing

1: Compute MEMS relative velocity array

\dot{z}

Array =

\dot{x}

Array −

\dot{y}

Array
2: Generate Expected Output array of rectangle classifier Y_Rof length T
3: Generate Expected Output array of triangle classifier Y_Tof length T
4: Initialize X
5: Input δ
6: Apply low-pass filter to

\dot{z}

Array
7: Compute zArray by integrating

\dot{z}

Array
8: Shift zArray by δ/T_s elements
9: Downsample with sample rate θ: zArray → zArrayDownSampled
10: for i = 1,2,...,T do
11: Fill X[i,ALL] = ith chunk of N elements of zArrayDownSampled
12: end for
13: Input TrainSamples
14: Generate X_Train = X[1 : TrainSamples,ALL]
15: Generate Y_R,Train = Y_R[1 : TrainSamples]
16: Generate Y_T,Train = Y_T [1 : TrainSamples]
17: Generate X_Test = X[TrainSamples + 1 : T,ALL]
18: Generate Y_R,Test = Y_R[TrainSamples + 1 : T]
19: Generate Y_T,Test = Y_T [TrainSamples + 1 : T]
20: Generate trained weight of rectangle classifier W_oR=

{(X T r a i n^{T} X T r a i n)}^{- 1}

∗

(X T r a i n^{- 1} Y R, T r a i n)

21: Generate test set results of rectangle classifier

Y_{R}^{*}

= X_TestW_oR
22: Generate trained weight of triangle classifier W_oT =

{(X T r a i n^{T} X T r a i n)}^{- 1}

∗

(X T r a i n^{- 1} Y T, T r a i n)

23: Generate test set results

Y_{T}^{*}

= X_TestW_oT
24: Compute classification accuracy Success Rat