Table 1 Comparison of vision-based and sensor-based emotion recognition with our work.
From: RF sensing enabled tracking of human facial expressions using machine learning algorithms
References | Technology used | Activity used | AI model | Accuracy (%) |
|---|---|---|---|---|
Vision-based | Facial expression | ResNet-50 | 95.39 ± 1.41 | |
Vision-based | Facial expression | DNN | 85.57 | |
Vision-based | Facial expression | 3D-CNN and ConvLSTM | 98.83 | |
Vision-based | Facial expression | DBN | 96.25 | |
Sensor-based | Facial muscle movements | Cross-domain transfer learning | 80.75 | |
Sensor-based | Emotion recognition | Random Forest | 60–70% | |
Sensor-based | Respiration and heart rate signals with emotion recognition | CNN and GRU | 84.5% and 74.25% | |
Sensor-based | Emotion recognition | Neural network | 80.59 | |
Our | Sensor-based | Emotion recognition | LSTM | 91.0% |
