Table 2 Ablation study on module contributions within the CHMMConvScaleNet framework.

From: CHMMConvScaleNet: a hybrid convolutional neural network and continuous hidden Markov model with multi-scale features for sleep posture detection

Method

Time scale

Conv1D

Conv2D

CHMM

Acc

Rec

Pre

F1 score

M 1

 

\(\checkmark\)

  

0.7081

0.6679

0.6812

0.6724

M 2

  

\(\checkmark\)

 

0.7979

0.7818

0.7686

0.7744

M 3

 

\(\checkmark\)

\(\checkmark\)

 

0.8653

0.8395

0.8460

0.8421

M 4

 

\(\checkmark\)

 

\(\checkmark\)

0.7560

0.7153

0.7182

0.7142

M 5

  

\(\checkmark\)

\(\checkmark\)

0.8558

0.8256

0.8418

0.8329

M 6

 

\(\checkmark\)

\(\checkmark\)

\(\checkmark\)

0.9119

0.8965

0.9015

0.8989

M 7

\(\checkmark\)

\(\checkmark\)

\(\checkmark\)

 

0.8989

0.8821

0.8917

0.8865

M 8

\(\checkmark\)

\(\checkmark\)

\(\checkmark\)

\(\checkmark\)

0.9341

0.9291

0.9187

0.9235