Table 1.

Input vectors used for modeling. Twenty gait cycles (GCs) per subject were extracted. To reduce the number of irrelevant features of V_simple and V_tsfresh, the fresh-algorithm (FeatuRe Extraction based on Scalable Hypothesis tests) [18] was applied as a filter. ROM = range of motion

Abbreviation	Description	Size (GC × Feature)
V_waves	Concatenated time-normalized GC for the measured variables.	940 × 2100
V_simple	Calculated features based on simple descriptive statistics which are commonly mentioned in the literature [35,36]. Maxima, minima, and ROM for every variable as well as the difference between affected and unaffected sides for the respective variables were calculated.	940 × 74
V_tsfresh	Automated feature extraction with the tsfresh algorithm [18].	940 × 8349