Table 3.
Studies that use the sliding window approach (Part 2).
| Publication (Number of Subjects) | Activities (Number of Activities) | Accelerometer Placements (Number of Accelerometers) | Inter-Subject Classification Accuracy | Window Sizes (in seconds) |
|---|---|---|---|---|
| Khan et al. (2010) (6 subjects) [53] | Walking, upstairs, downstairs, running, sitting (5) | Smartphone in 5 different pocket locations (shirt's top, jeans' rear/front-left/front-right, coat's inner) (1) | ANN-OF (46%) ANN-LDA (60%) ANN-KDA (96%) | 2 |
| Marx (2010) (1 subject) [31] | Ball interactions (throwing, shaking, jerking sideways, holding very still) (4) | Embedded in iBall (1) | Heuristic (90%–95%) | 0.666 |
| Sun et al. (2010) (7 subjects) [54] | Walking, running, stationary, upstairs, downstairs, driving, bicycling (7) | Front/rear pockets on the trousers, front pockets on the coat (6) | SVM (93% with acceleration magnitude in 4 s; 92% without acceleration magnitude in 5 s) | 1, 2, 3, 4, 5, 6 |
| Atallah et al. (2011) (11 subjects) [55] | Reading, socializing, vacuuming and more (15) | Chest, arm, wrist, waist, knee, ankle, right ear (7) | kNN with k = 5 (∼56%) and k = 7 (∼64%), NB with Gaussian priors (∼61%) | 5 |
| Gjoreski and Gams (2011) (11 subjects) [56] | Standing, sitting, lying, sitting on the ground, on all fours, going down, standing up (7) | Chest, left thigh, right ankle (3) | Random Forest (93% only with chest; 96% adding left thigh; 98% with all accelerometers) | 1 |
| Jiang et al. (2011) (10 subjects) [57] | Walking, jogging, weight lifting, cycling, rowing and more (10) | Both forearms and shanks (4) | SVM ideal (95.1%) SVM with errors (75.2%) SVM without orientation errors (91.2%) SVM without errors (91.9%) | 6.4 |
| Kwapisz et al. (2011) (29 subjects) [58] | Walking, jogging, upstairs, downstairs and more (6) | Smartphone (1) | DT (85.1%) LR (78.1%) MLP (91.7%) | 10 |
| Lee and Cho (2011) (3 subjects) [59] | 3 actions (walking, standing, climbing stairs) + 3 activities (shopping, moving by walk, taking bus) | Smartphone in the hand (1) | HHMM (84%) HMM (65%) ANN (65%) | 5 |
| Siirtola and Röning (2012) (8 subjects) [60] | Walking, running, cycling, sitting/standing, driving a car (5) | Smartphone in trousers' front pocket (1) | Offline (QDA (95.4%) kNN (94.5%)) Real-Time with Nokia (QDA (95.8%) kNN (93.9%)) Real-time with Samsung Galaxy (QDA (96.5%)) | 7.5 |
| Wang et al. (2012) (8 subjects) [61] | Walking, jogging, upstairs, downstairs (4) | Smartphone (1) | GMM (91.2%) J48 (88.8%) LR (93.3%) | 0.5, 0.8 |
| Hemalatha and Vaidehi (2013) (5 subjects) [62] | Walking, sitting/standing, lying, falling (4) | Chest (1) | FBPAC (92%) | 10 |
| Mannini et al. (2013) (33 subjects) [26] | 4 broad activity classes (ambulation, cycling, sedentary and other), daily activities (26) | Wrist or ankle (1) | SVM (84.7% with wrist, 95% with ankle) for 12.8 s | 2, 4, 12.8 |
| Nam and Park (2013) (3 subjects) [63] | Walking, toddling, crawling, wiggling, rolling and more (11) | Waist (1) | NB (81%) BN (87%) DT (75%) SVM (95%) kNN (96.2%) J48 (94.7%) MLP (96.3%) LR (93.2%) | ∼2.7 |
| Nam and Park (2013) (11 subjects) [64] | Walking, toddling, crawling, wiggling, rolling and more (10) | Waist (1) | NB (73%) BN (84.8%) DT (74%) SVM (86.2%) kNN (84.1%) J48 (88.3%) MLP (84.8%) LR (86.9%) | ∼2.7 |
| Zheng et al. (2013) (18/53/7 subjects) [65] | 3 datasets: Walking, running, dancing and more (7) in 1st & 2nd / Walking, jogging, skipping and more (6) in 3rd | Wrist (1 in 1st) Hip (1 in 2nd) Waist pocket (1 in 3rd) | SWEM-SVM (94%/90%/82%) SVM (93%/89%/79%) ANN (91%/78%/74%) | 10 |