TABLE 1.
Comparison of the proposed single-wrist approach FoG detection with other studies exploiting one or more sensors to capture movement.
| Sensors | Sensors placement | Realtime | Results | |
|---|---|---|---|---|
| Without sensor in upper limb | ||||
| Bachlin et al. (2009) | 3 | Ankle, thigh and lower back | Yes | 73.1% sensitivity |
| 81.6% specificity | ||||
| Jovanov et al. (2009) | 2 | Both sensors on belt or knee or ankle or shoe | Yes | unknown |
| Niazmand et al. (2011) | 5 | Shank and belt | No | 88.3% sensitivity |
| 85.3% specificity | ||||
| Mazilu et al. (2015a) | 2 | Ankles | Yes | 97% hit rate |
| Palmerini et al. (2017) | 3 | Shins, lower back | Yes | 83% sensitivity |
| 67% specificity | ||||
| Camps et al. (2018) | 1 | Waist | Yes | 92.6% sensitivity |
| 88.7% specificity | ||||
| With sensor in upper limb | ||||
| Cole et al. (2011) | 3 | Shank, thigh and arm | No | 83% sensitivity |
| 97% specificity | ||||
| Tripoliti et al. (2013) | 6 | Wrists, ankles, waist and chest | No | 81.9% sensitivity |
| 98.7% specificity | ||||
| Mazilu et al. (2015b) | 2 | Both wrists | Yes | 90% sensitivity |
| 83% specificity | ||||
| DeepFoG | 1 | Wrist | Yes | 83% sensitivity |
| 88% specificity | ||||