SlidingWindowGaitEstimation

Example code of
Simultaneous Locomotion Mode Classification and Continuous Gait Phase Estimation for Transtibial Prostheses
Ryan R. Posh* $^1$, Shenggao Li* $^2$, Patrick M. Wensing $^2$

Background

Recognizing and identifying human locomotion is a critical step to ensuring fluent control of wearable robots, such as transtibial prostheses. In particular, classifying the intended locomotion mode and estimating the gait phase are key. In this work, a novel, interpretable, and computationally efficient algorithm is presented for simultaneously predicting locomotion mode and gait phase. Using able-bodied (AB) and transtibial prosthesis (PR) data, seven locomotion modes are tested including slow, medium, and fast level walking (0.6, 0.8, and 1.0 m/s), ramp ascent/descent (5 degrees), and stair ascent/descent (20 cm height). Overall classification accuracy was 99.1% and 99.3% for the AB and PR conditions, respectively. The average gait phase error was less than 4% across all data. Exploiting the structure of the historical data, coumputational efficiency reached 2.91 $\mu$ s per time step. The time complexity of this algorithm scales as $\mathcal{O}(N\cdot M)$ with the number of locomotion modes $M$ and samples per gait cycle $N$. This efficiency and high accuracy could accommodate a much larger set of locomotion modes ($\sim$ 700 on OpenSource Leg Prosthesis) to handle the wide range of activities pursued by individuals during daily living.

Dependencies and Requirements

• OS: Crossplatform, but tested on MacOSX or Linux (x86/x64 and/or ARM)
• C++ Compiler: g++ (recommended) or clang++

  g++ compiled code in our test is 10~20% faster than the clang++ compiled one

• Eigen - (Optional) for implementation using Eigen library.

  Eigen is a header library for linear algebra. It provides explicit vectorization for matrix operations and is better optimized. Therefore, it reduce the computation time of naive/plain approach to a fraction of pointer implementation. The setup script would tell makefile to skip Eigen implementation if no Eigen library is found.

• MATLAB - For converting, processing, and plotting data.

  We use PopIn to read and converting .yaml format data in MATLAB. The setup script can automatically download it to directory MATLAB/PopIn

Getting Started

Install

Download this project

git clone https://github.com/ROAM-Lab-ND/SlidingWindowGaitEstimation.git
cd SlidingWindowGaitEstimation

Run the setup script and follow the instruction to configure compiler and Eigen.

./setup.sh

or specify a compiler and Eigen path (you must specify compiler then specify Eigen)

./setup.sh [COMPILER] [/PATH/TO/EIGEN]

Available Makefile commands

1. make clean to remove all compiling and executable files of this project.
2. make all to compile all.

Run Example

• Run the script to test the C++ code:

./runtest.sh

The computation time would be output as

Start: Trick (moving window) - pointer implementation
Total Time (us): 10073
Time Per Data (us): 3.35767

• Or run the MATLAB script MATLAB/SSE_without_savings.m to run the MATLAB implementation.
• Script MATLAB/DebugTest/CPPClassificationTest.m will compare the result between C++ and MATLAB implementation

Flag Definitions

We have four different implementations with two set of flags: Trick/Plain and Homemade/Eigen. Suppose we have $M$ activities and each activity has $N$ data.

• Trick - using sliding window trick, time complexity $\mathcal{O}(N\cdot M)$, space complexity $\mathcal{O}(M\cdot N^2)$.
• Plain - (a.k.a., Naive in paper) no computation save, compute SSE from scratch, time complexity $\mathcal{O}(M\cdot N^2)$, space complexity $\mathcal{O}(N\cdot M)$.
• Homemade - (i.e., pointer) store and compute array/matrix with pointer, probably less overhead.
• Eigen - store array/matrix with Eigen array type, and compute with Eigen function, better vectorization due to explicit AVX/SIMD call, cleaner expression.

C++ Implementation Usage

Direct call C++ program via command line.

./build/SSExxxxx /DIR/OF/DATABASE/ /PATH/TO/EXPERIMENT/DATA/ INDEX_START INDEX_END /PATH/TO/RESULT/FILE

For example:

./build/SSEtrick ./Data/DataBase_CSV ./Data/Experiment_CSV/Multispeed_Walk_AB.csv 4000 7000 ./CPPResult/Test_Trick.csv

Citation

{
...
}

Contact

$^{1}$ University of Michigan, Ann Arbor, MI, US. poshr@umich.edu
$^{2}$ University of Notre Dame, South Bend, IN, US. {sli25,pwensing}@nd.edu