Table 4.
Performances of reported sleep staging using the signals from nearable devices
| Type | Signals / Models | # of Stages |
# of Rec. |
Performance | Ref. | ||
|---|---|---|---|---|---|---|---|
| ACC (%) | Kappa | ||||||
| Radar |
Sig_hba., Sig_ra., Mov. / LSTM |
4 | 51 | 82.6 | 0.73 | [45] | |
|
Sig_hba., Sig_ra., Mov. / K-Nearest Neighbor |
4 | 13 | 81.0 | - | [46] | ||
| RF spectrogram / Conditional Adversarial Discriminator | 4 | 100 | 79.8 | 0.70 | [47] | ||
| Resp. Mov. / LSTM | 4 | 71 | 76.0 | 0.63 | [48] | ||
| - | 4 | 40 | 70.0 | 0.53 | [49] | ||
| Microphone | Resp.-related sound | Feedforward neural network | 3 | 250 | 86.9 | 0.69 | [50] |
|
CNN + LSTM +Transformer |
4 | 1481 | - | 0.52 | [51] | ||
| Film / bed installed | BCG, Resp. Mov. / LSTM | 4 | 60 | 73.9 | 0.55 | [52] | |
| BCG, Resp. Mov. / DNN | 2 | 45 | 86.0 | 0.45 | [53] | ||
| Resp. Mov / Rule-based model | 4 | 25 | 70.9 | 0.48 | [54] | ||
| BCG, Resp. Mov. / - | 4 | 102 | 79.0 | 0.68 | [55] | ||
| BCG, Resp. Mov. / - | 4 | 85 | 64.5 | 0.46 | [56] | ||
# of Rec. number of data points, ACC. accuracy, Ref. reference, Sig_hba. signal, reflecting the heartbeat activity from the radar signal, Sig_ra. signal reflecting the respiratory activity from the radar signal, Mov. Movement, RF radio frequency, Resp. Respiration, BCG ballistocardiogram, LSTM Long-Short Term Memory, CNN convolutional neural network, DNN deep neural network