Skip to main content
. 2024 Jan 22;5:163–172. doi: 10.1109/OJEMB.2024.3356791

TABLE I. Overview of the Core Algorithms Provided in the Gaitmap Package.

Processing Step Algorithm Description
Preprocessing align_dataset_to_gravity Alignment of IMU signal with gravity.
align_heading_of_sensors Align the heading to the movement direction using PCA on the gyroscope signal and an integration of the acceleration signal (unpublished, inspired by [40]).
Gait Detection UllrichGaitSequenceDetection Detection of gait using harmonic frequencies from raw IMU signals [41]*.
Stride Segmentation BarthDtw Template matching using dynamic time warping (DTW) modified based on [42]*.
ConstrainedBarthDtw BarthDtw with additional local warping constraints. Inspired by local weightings explained in [43].
HmmStrideSegmentation Multi-activity Hidden-Markov-Model (HMM) based on [44]* implemented using [37].
Event Detection RamppEventDetection Initial and terminal foot contact detection based on signal features [39]*.
FilteredRamppEventDetection RamppEventDetection with additional lowpass filtering for specific calculation steps.
HerzerEventDetection Initial and terminal foot contact based on signal features tuned for stair walking [45]*. Combines concepts from [39]* and [46].
Zero-Velocity Detection AredZuptDetector Window based thresholding on gyroscope norm to detect static regions [47].
NormZuptDetector Window based thresholding to detect static regions. Generalized version of ARED from [47].
ShoeZuptDetector Window based thresholding on combined gyroscope and accelerometer norm to detect static regions [47].
Trajectory Reconstruction RtsKalman Zero-Velocity aided Error Tracking Kalman Filter with smoothing to estimate the foot trajectory from IMU signals. Based on [48], [49] with implementation details from [50].
MadwickRtsKalman Variation of the RtsKalman method using the Madgwick algorithms for orientation estimation (unpublished, based on [24], [48], [49], [50]).
SimpleGyroIntegration Orientation estimation based on integration of the gyroscope signal.
MadgwickAHRS Sensor fusion algorithm for orientation estimation [24].
ForwardBackwardIntegration Position estimation using double-integration of the acceleration with dedrifting through forward-backwards integration [51] and [9]*.
PieceWiseLinearDedriftedIntegration Position estimation using double-integration of the acceleration with linear drift model (unpublished, inspired by [52]).
Parameter Estimation TemporalParameterCalculation Calculation of temporal parameters (e.g. stride time) based on the detected events.
SpatialParameterCalculation Calculation of spatial parameters (e.g. stride length) based on the calculated trajectory [39], [53]*.

The Name in the Algorithm Column Corresponds to the class/function Name Within the Package. The Last Column Lists the Publications That the Algorithm is Based On. Algorithms Marked With an Asterisk (*) are Publications That Were Published by the Authors of This Manuscript Directly or the Associated Research Group. Besides the Core Algorithms, the Package Provides a Number of Helper Methods, Which are Not Listed Here. For a Full List See the Official Documentation Page (gaitmap.readthedocs.io). The Categories are Listed in the Order They are Typically Used in a Gait Analysis Pipeline.