|
| System | Users | Signal/preprocessing | Experimental scene | Purpose/classification | Performance |
|
| WFID [62] | Corridor: 9
Laboratory: 6 | Amplitude/PCA | Corridor, laboratory (155 m2) | Identity recognition/SVM | 6 people: 93.1%
9 people: 91.9% |
|
| WiFi-ID [65] | 20 | Amplitude/Butterworth filter, CWT | Corridor | Identity recognition/SAC | 2 to 6 people: 93% to 77% |
|
| FreeSense [51] | 9 | Amplitude/PCA, DWT, low-pass filter | A smart home environment () | Identity recognition/KNN, DTW | 2 to 6 people: 94.5% to 88.9% |
|
| WiWho [58] | 20 | Amplitude/multipath removal, band-pass filter | Three indoor environments | Identity recognition/decision tree | 2 to 6 people: 92% to 80% |
|
| WifiU [69] | 50 | Amplitude/PCA, STFT | A typical lab (50 m2) | Identity recognition/LibSVM with RBF kernel | 50 people: top 1: 79.28%
Top 2: 89.52%
Top 3: 93.05% |
|
| Jakkala et al. [71] | 30 | Amplitude/Hanning window | An office | Identity recognition/DCNN | 30 people: |
|
| AutoID [72] | 20 | Amplitude/DWT | A conference room (), an office zone (), a bedroom apartment () | Identity recognition/convex clustered concurrent Shapelet learning | 20 people: 91% |
|
| Nipu et al. [73] | 5 | Amplitude/Butterworth low-pass filter | An opened room | Identity recognition/decision tree, random Forest | 2 to 5 people: 95% to 84% (decision tree), 97.5% to 78% (random Forest) |
|
| Neural Wave [74] | 24 | Amplitude and phase/WT, IWT, and PCA | A typical indoor laboratory | Identity recognition/1-D ConvNet, called RadioNet (23 layers) | 24 people: |
|
| Wide [75] | 10 | Amplitude/PCA | A laboratory | Identity recognition/SVM | Open scene: 98.7%
No interference: 100% eliminate scene disturbances: 99.7% |
|
| Nkabiti et al. [76] | 7 | Amplitude and phase/Chebyshev filter | Dormitory room () and hallway | Identity recognition/LSTM-RNN | Dormitory: 95.5%
Hallway: 96.3% |
|
| Wii [59] | 8 | Amplitude and phase/PCA, CWT, low-pass filter | A meeting room () | Identity recognition stranger identification/SVM, GMM | 2 to 6 people: 98.7% to 90.9%
Stranger identification accuracy: about 93% of 2 strangers |
|
| CareFi [70] | 16 | Amplitude/low-pass filter, PCA, STFT | A typical meeting room (), an apartment () | Stranger identification/SVM | Intruder detection: more than 87.2% |
|
| RDFID [77] | 4 | Amplitude and phase/PCA, CWT | A meeting room (), a living room (), a large office () | Stranger identification/SVM, GMM | Stranger identification accuracy: around 79%
FN: around 2%
FP: around 2% |
|
| HumanFi [64] | 24 | Amplitude and phase/Butterworth filter, a method proposed in [78] | The doorway of an office, the middle of an office | Identity recognition/LSTM | 24 people: 96% |
|
| Deep-WiID [52] | 15 | Amplitude | Hall, lab | Identity recognition/GRU, average pooling | 2 to 6 people: 99.7% to 97.7%
15 people: 92.5% |
|
| CSIID [79] | 6 | Amplitude | An indoor environment | Identity recognition/convolution layer, LSTM | 2 to 6 people: 97.4% to 94.8% |
| WiDIGR [80] | 60 | Amplitude/band-pass filter, PCA | A laboratory, an empty room, and an apartment | Identity recognition/SVM | 3 to 6 people: 92.83% to 78.28% |
|
| Gate-ID [81] | 20 | Amplitude/silence removal algorithm [65] | A room | Identity recognition/ResNet and bi-LSTM | 6 to 20 people: 90.7% to 75.7% |
|
