Skip to main content

Table 4 The binary accuracy and F1 scores for cross-testing single and combined source datasets on the PD dataset

From: Deep neural networks for wearable sensor-based activity recognition in Parkinson’s disease: investigating generalizability and model complexity

 

Base Model

O’Halloran [13]

Wan [14]

Burns [15]

Chen/Jordao [10, 16]

Kalouris [17]

Yatbaz [18]

Rueda [19]

\(\begin{array}{l}\hbox {Source: P2 Orig} \\ \hbox {Target: PD Reorient} \end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}\hbox {Source: MH Orig} \\ \hbox {Target: PD Reorient} \end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}62.38\\ (2.88)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}\hbox {Source: P2 Aug} \\ \hbox {Target: PD Reorient}\end{array}\)

\(\begin{array}{l}65.23\\ (33.66)\end{array}\)

\(\begin{array}{l}62.86\\ (26.27)\end{array}\)

\(\begin{array}{l}64.32\\ (28.63)\end{array}\)

\(\begin{array}{l}66.46\\ (0)\end{array}\)

\(\begin{array}{l}66.07\\ (16.50)\end{array}\)

\(\begin{array}{l}63.89\\ (22.48)\end{array}\)

\(\begin{array}{l}65.78\\ (17.81)\end{array}\)

\(\begin{array}{l}61.26\\ (17.82)\end{array}\)

\(\begin{array}{l}\hbox {Source: MH Aug} \\ \hbox {Target: PD Reorient}\end{array}\)

\(\begin{array}{l}72.18\\ (30.78)\end{array}\)

\(\begin{array}{l}70.64\\ (36.72)\end{array}\)

\(\begin{array}{l}71.17\\ (35.55)\end{array}\)

\(\begin{array}{l}66.01\\ (19.14)\end{array}\)

\(\begin{array}{l}67.45\\ (11.82)\end{array}\)

\(\begin{array}{l}64.34\\ (28.15)\end{array}\)

\(\begin{array}{l}68.53\\ (25.81)\end{array}\)

\(\begin{array}{l}68.63\\ (34.66)\end{array}\)

\(\begin{array}{l}\hbox {Source: P2 Aug AND} \\ \hbox {MH Reorient Aug} \\ \hbox {Target: PD Reorient}\end{array}\)

\(\begin{array}{l}55.56\\ (35.80)\end{array}\)

\(\begin{array}{l}71.25\\ (44.96)\end{array}\)

\(\begin{array}{l}73.74\\ (49.04)\end{array}\)

\(\begin{array}{l}66.37\\ (1.30)\end{array}\)

\(\begin{array}{l}63.85\\ (16.51)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}66.67\\ (0)\end{array}\)

\(\begin{array}{l}67.37\\ (25.13)\end{array}\)

  1. The average values are shown, with F1 scores presented in parentheses. The notation for data augmentation applies to both the MHEALTH and PAMAP2 datasets. ‘PD Reorient’ indicates PD data with sensor axes reoriented to match the source data, or to PAMAP2 in the case of the combined source datasets