Skip to main content

Table 3 Overview of accelerometry-based methods

From: Wearable accelerometry-based technology capable of assessing functional activities in neurological populations in community settings: a systematic review

Authors

Population

Method

Validity

Quality

Quantity

Activity

Lau et al [34]

Stroke

SVM, MLP, RBF

Leave-one-subject-out method

-

-

Walking

Barth et al [31]

PD

Boosting with decision stump as weak learner, LDA, and SVM with linear and RBF kernel

Leave-one-subject-out method

x

x

Walking, foot circling, and heel-toe tapping

Cancela et al [32]

PD

k NN, Parzen, Parzen density, binary decision tree, Bpxnc train NN by back-propagation, and SVM

Cross-validation

x

-

Daily activities (i.e. walking, lying, sitting, drinking a glass of water, opening and closing a door)

Salarian et al [32]

PD

Logic Regression model with Mamdani fuzzy rule-based classifier

Cross-validation

-

x

sit-to-stand and stand-to-sit

Zwartjes et al [45]

PD

Decision tree

Leave-one-subject-out method

x

-

lying, sitting, standing, and walking

Yang et al [43]

PD

Autocorrelation method

Video recordings

-

x

Walking

Motoi et al [39]

Stroke

Sagittal angle changes

 

-

-

Walking and sit-to-stand

Moore et al [38]

PD

Mathematical step-length algorithm

Pen techniques and video recordings

x

x

Walking

Dobkin et al [33]

Stroke

Naive Bayes classifier in combination with Gaussian discretization followed by a maximum likelihood estimation

Stopwatch

-

x

Walking

  1. Abbreviations: LDA Linear Discriminant Analysis, SVM Support Vector Machines, RBF radial basis function neural network, K-NN K-nearest neighbour, NN Neural Network, MLP multi-layer perception; Quality, methods assessing severity levels, Quantity, methods able to distinguish healthy from non-healthy subjects.