Table 10.
A summary of detection algorithms employed on RIP sensors.
| Ref | No of RIPBand | Pre-processing | De-noising | Artifact Removal | Feature Extracted | Classifier Employed |
Signal Classification | Validation | Study Type | Performance Matrices |
|---|---|---|---|---|---|---|---|---|---|---|
| [89] | 2 (Thoracic TC, Abdominal AB) |
1. Tidal Volume and Airflow measurement from TC, AB signals 2. Signal Normalization to the range of -1 and 1 |
- | An ideal band pass filter, fc = 0.0001–10 Hz | - | Simple Peak-Valley Detection |
4 activities (resting, reading aloud, food intake and smoking) |
Train- 5 fold cross-val; Test-LOOS |
Lab, 20 subject | Accuracy: Resting-0.96, Reading-0.89, Food intake-0.91, Smoking-0.89 |
| [90] | Average Gaussian filter of 25 points | Z-norm 16 features Using Window 0.5s, 50% overlap |
Left-to-right hidden Markov models | 5 activities (sedentary, walking, eating, talking, and cigarette smoking) | LOOS | Lab, 20 subject | Precision 0.60, Recall 0.67 F1-score 0.62 |
|||
| [86] | 1 (Thoracic TC) |
- | - | 17 features from each 30s window | Supervised and semi-supervised support vector | Puff or non-puff | LOOS | Lab, 10 subject | Accuracy 0.91 |