code
These are some of my coding projects. Most of it was during my PhD work at Pitt’s Sensorimotor Learning Lab, and are coded for Matlab (although I am now coding mostly in Python).
matlab-linsys
Linear dynamic systems toolbox for Matlab. Includes tools for identification (EM and subspace methods), selection (BIC, AIC, cross-validation), visualization, and comparison of linear systems. It implements the Kalman filter and Kalman smoother (in covariance, information and square root formulations), with emphasis on filtering speed. Offers an object-oriented wrapper of most of its functionality to facilitate its adoption. Repository
monoLS
Data smoothing through monotonic curve fitting for Matlab and Octave.
MonoLS finds the optimal monotonic (i.e. non-negative or non-positive derivative) curve fit to some dataset of (x,y) points. Higher-order derivatives can also be constrained to be of constant sign (or 0), but the same sign has to be used for all odd-order derivatives and for all even-order derivatives. A regularization term is used to avoid undue deference to datapoints near the extremes of the x range. By default, monoLS uses mean-squared errors so it is the least-square solution. In this case, the problem can be framed as a non-negative least squares problem (which is always convex).
Repository
robustCov
A simple, fast, heuristic, robust covariance matrix estimator to deal with data with outliers (MatlaB). Repository
pi-tools
Collection of Matlab utilities to make my life easier. Includes functions for visualization, non-parametric statistical tests, and filtering. Repository
labTools
A suite of Matlab utilities to load, clean, process, visualize and analize experimental gait data. Used in all past and present studies by Pitt’s Sensorimotor Learning Lab. Experimental data is organized and stored in custom-made hierarchical object-oriented structures, and most functionality is implemented as class-specific methods for these objects. Repository
HMRLHardware
Interfacing functions to communicate with Pitt SML Lab’s (formerly HMRL) hardware and software from Matlab to have a single point of control for experiments. Hardware includes a Bertec split-belt treadmill, and software includes Vicon’s Nexus. Repository
splitbelt-EMG-adaptation
Data and code (Matlab) for the research project published in Corrective Muscle Activity Reveals Subject-Specific Sensorimotor Recalibration Repository