. 2023 Jul 24;10:1160373. doi: 10.3389/fmed.2023.1160373

Table 1.

Fetal movement recognition algorithm.

Algorithm

Input: Pre-recognition signal Xi, i = 1, 2, …, 6.
Feature dictionary Dk, k = 1,2.
Sparsity coefficient T₀.

Output: Recognition category Yi, i = 1, 2, …, 6.
Initialize the parameters.
for i = 1:6
for k = 1:2
while t < T₀
Iteration number t = 1, reconstructed residual r₀ = X_i, set Λ₀ = Ø, D=D_k.
Find the index λ_t where the reconstructed residual r_t-1 matches the feature dictionary D, i.e.,

λ_{t} = a r g m a x_{j \in Λ_{t - 1}} | r_{t - 1}, d_{j} |

.
Update the set of indexes Λ_t = Λ_t-1∪λ_t, reconstructing the dictionary D_t = [D_t-1, d_λt].
Compute using least squares to obtain

_{\hat{x} t} = \arg \min_{\hat{\hat{x}}} ∥ y - D \hat{x} ∥_{2}^{2}

.
Update the reconstructed residuals

r_{t} = y - D_{t} {\hat{x}}_{t}

.
t = t + 1.
end while
Cache the sparse residuals e_k = r_t.
end for
if e₁ < e₂, identify the signal as fetal Y_i = 1,
else, identify the signal as non-fetal Y_i = 0.
end for