Data collected by four channel LDC1614, at different RCOUNT, from a moving resonator attached to the robotic eye

We have collected data using a four-channel inductance-to-digital converter (LDC1614) to measure the motion of a robotic eye moving at discrete steps. A resonator tuned to nearly the same frequency as the detectors was embedded in a contact lens and placed on the robotic eye. The movement of the resonator-embedded contact lens placed on the robotic eye results in changes in the measured inductance of the detector. The LDC1614 registers are fixed except the RCOUNT, which varies. The folder names indicate the RCOUNT register value. The preprocessed data are the raw data that were processed through a multistage, zero-phase preprocessing pipeline to each channel prior to modeling, in the following order: Butterworth \u2192 Hampel \u2192 Savitzky\u2013Golay \u2192 robust rescaling. All filtering was performed in zero phase using forward\u2013backward application to avoid phase distortion. A 54th-order digital Butterworth filter with a cutoff frequency of 2.375 Hz (sampling rate \u2248 55.556 Hz, Nyquist = 27.778 Hz; normalized cutoff \u2248 0.0855) was applied using a 2nd-order cascade and zero-phase filtering (i.e., sosfiltfilt), with an effective padding of about 3\u00d7order = 162 samples to minimize edge transients. To suppress impulsive artifacts that pass through the linear low pass, we used a Hampel filter with window = 21 samples (\u2248 0.38 s) and \u03c3 = 3, replacing detected outliers with the local median. We then applied a Savitzky\u2013Golay smoother with a window of 21 samples (odd, > order) and a 4th-order polynomial to reduce high-frequency jitter while preserving local curvature. Finally, channels were normalized via median\/IQR scaling: