2.2 Preprocessing

Data three-dimensional (3D) IMU measurements are extracted for every PD patient from the CuPiD dataset. In particular, for each PD patient, the corresponding N-sample accelerometer and gyroscope data (sampled at fs=128 Hz), i.e., ({a _x: [ax[1], ax[2],…, ax[N]]^T}, {a _y: [ay[1], ay[2],…, ay[N]]^T}, {a _z: [az[1], az[2] …, az[N]]^T}) and ({g _x: [gx[1], gx[2],…, gx[N]]^T}, {g _y: [gy[1], gy[2],…, gy[N]]^T}, {g _z: [gz[1], gz[2] …, gz[N]]^T}) respectively, were constructed. The data are processed and formatted as a N×6 matrix for every PD patient, i.e., Sj = [a _x,a _y,a _z,g _x,g _y,g _z], j=1,…,11 (PD patients).

Further, a three-second (Wl=3 s) sliding window with 0.25 s overlap is being selected for processing the Sj data. The duration of the window and the overlap have been chosen as a trade-off between the duration of FoG events and the time resolution needed to detect a differentiation in wrist movement (Mazilu et al., 2015a). Every element of Sj is being normalized with the maximum value per vector across all PD patients and rounded to 4 decimal digits for smaller bit storage.