Skip to main content

Table 6 Segmentation performance on the free-living dataset, grouped by the number of strides per walking bout. All values are given as mean ± std

From: Hidden Markov Model based stride segmentation on unsupervised free-living gait data in Parkinson’s disease patients

N Strides per bout Precision [%] Recall [%] F1-Score [%]
DTW HMM\(_{\mathrm{lab}}\) \( \hbox {HMM}_{\mathrm{free}\hbox {-}\mathrm{living}}\) DTW HMM\(_{\mathrm{lab}}\) \( \hbox {HMM}_{\mathrm{free}\hbox {-}\mathrm{living}}\) DTW HMM\(_{\mathrm{lab}}\) \( \hbox {HMM}_{\mathrm{free}\hbox {-}\mathrm{living}}\)
\(4 \le \hbox {N} \le 15\) \(69.2 \pm 10.2\) \(75.7 \pm 4.5\) \(75.5 \pm 4.5\) \(59.6 \pm 12.4\) \(92.7 \pm 3.7\) \(93.2 \pm 3.7\) \(63.5 \pm 10.2\) \(83.3 \pm 3.7\) \(83.4 \pm 3.8\)
\(15 < \hbox {N} \le 30\) \(85.5 \pm 8.9\) \(88.2 \pm 4.7\) \(88.7 \pm 4.7\) \(80.5 \pm 10.4\) \(94.1 \pm 3.3\) \(94.7 \pm 3.3\) \(82.8 \pm 9.1\) \(91.1 \pm 3.8\) \(91.6 \pm 3.5\)
\(30 < \hbox {N} \le 50\) \(90.8 \pm 5.1\) \(92.7 \pm 4.2\) \(92.6 \pm 4.2\) \(88.6 \pm 7.9\) \(95.0 \pm 3.7\) \(95.5 \pm 3.7\) \(89.6 \pm 6.1\) \(93.8 \pm 3.7\) \(94.0 \pm 3.6\)
\(50 < \hbox {N} \le 100\) \(93.9 \pm 3.8\) \(95.7 \pm 2.8\) \(95.9 \pm 2.8\) \(91.8 \pm 5.7\) \(96.2 \pm 2.5\) \(96.5 \pm 2.5\) \(92.8 \pm 4.6\) \(95.9 \pm 2.4\) \(96.2 \pm 2.0\)
\(100 < \hbox {N} \le 200\) \(95.1 \pm 3.6\) \(96.3 \pm 3.0\) \(96.4 \pm 3.0\) \(92.8 \pm 5.1\) \(95.8 \pm 2.9\) \(96.1 \pm 2.9\) \(93.9 \pm 4.2\) \(96.0 \pm 2.9\) \(96.2 \pm 3.0\)
\(\hbox {N} > 200\) \(97.9 \pm 2.2\) \(98.4 \pm 1.8\) \(98.7 \pm 1.8\) \(96.4 \pm 4.0\) \(97.4 \pm 2.6\) \(97.8 \pm 2.6\) \(97.2 \pm 3.0\) \(97.9 \pm 2.1\) \(98.2 \pm 1.7\)