Table 1 Comparison for brain signal activity analysis.
Classification method | Description | Quality metric |
|---|---|---|
Cold and hot voxels | A technique for detecting weariness based on thermal imaging yawning detection. Using the hot and cold voxels, yawning was detected. | Accuracy Cold voxels: 71% Hot voxels: 87% |
SVM and KNN | Examined the driver’s breathing using facial thermal imaging to determine whether or not they were sleepy. | Accuracy: SVM:90% KNN = 83% Sensitivity SVM 92% KNN 82% Precision SVM 91%, KNN 90% |
Classification based on the period of eye closure | The device detected whether an eye was open or closed based on the concavity of the eyelid’s curvature. The eye closure duration was then used to determine tiredness. | Accuracy Dataset 1:90% Dataset 2: 70% Dataset 3: 95% |
Proposed optical correlation with deformed filter | To precisely determine the location of the eyes in the vander lugt correlator’s fourier plane, use the optical vander lugt correlator. | Different accuracy for different datasets |
Multilayer perception, RF and SVM | Tracked how long eyes blinked during video broadcasts to use the EAR as a sleepiness indicator. In general, the SVM performed the best. | Accuracy SVM: 94.9% |
KNN, SVM, logistic regression and ANN | A nominstrusive method using state tracking of the face and eyes. The KNN and ANN models were found to be the best ones in the end. | Accuracy: KNN 72.25% ANN 71.61% Sensitivity KNN 83.33% ANN 85.56% |
FD-NN, TL-VGG16 and TL-VGG19 | Created a real-time CNN-based system based on the region of ocular closure. Three networks were introduced for the classification of ocular closure. TL-VGG19, FDNN, and TLVGG16 | Accuracy: FD-NN 98.15% TLVGG16 95.45% TLVGG19 95% |
