Measurement Noise Recommendation for Efficient Kalman Filtering over a Large Amount of Sensor Data

. 2019 Mar 7;19(5):1168. doi: 10.3390/s19051168

Algorithm 3

f_{D A E} ()

: Measurement noise variance by denoising autoencoder.

Input:

Θ = {T_{1}, T_{2}, \dots, T_{m}}

: a training set of m sensor data;

S_{s e n s o r}

: a numeric sequence of input sensor data;
Output:
R: measurement noise variance of

S_{s e n s o r}

;
begin
(0) DAEmodel

: =

TrainModel

(Θ)

;
// Build a denoising autoencoder model in the preprocessing step.
(1)

S_{e s t i m a t e} : =

ApplyModel

(S_{s e n s o r}, DAEmodel)

;
(2)

R : =

CalculateVAR

(S_{s e n s o r}, S_{e s t i m a t e})

; // Calculate the variance between two sequences.
(3) return R;
end