Effectiveness of Landmark and Continuous Registrations in Reducing Inter- and Intrasubject Phase Variability

Zin, Muhammad Athif Mat; Rambely, Azmin Sham; Ariff, Noratiqah Mohd

doi:10.1109/access.2020.3038494

Cited by 3 publications

(2 citation statements)

References 36 publications

(72 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the FDA context, other, much more efficient strategies have been developed to reduce phase variability through different nonlinear registration procedures, as reviewed in [10]. Nonlinear time-scale normalization techniques include landmark registration [11], the sequence of states method [12], dynamic time warping [8,13], and curve registration based on correlation criteria [1,9,14] or on the Fisher-Rao Riemannian metric and square-root velocity function (SRVF) [15,16]. Function registration methods could virtually eliminate all phase variability so that the registered curves have the same shape, differing only by the different amplitudes [11,15].…”

Section: Introductionmentioning

confidence: 99%

“…Nonlinear time-scale normalization techniques include landmark registration [11], the sequence of states method [12], dynamic time warping [8,13], and curve registration based on correlation criteria [1,9,14] or on the Fisher-Rao Riemannian metric and square-root velocity function (SRVF) [15,16]. Function registration methods could virtually eliminate all phase variability so that the registered curves have the same shape, differing only by the different amplitudes [11,15]. This makes it possible to obtain good estimates of the average curves.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A New Method for Time Normalization Based on the Continuous Phase: Application to Neck Kinematics

et al. 2021

View full text Add to dashboard Cite

There is growing interest in analyzing human movement data for clinical, sport, and ergonomic applications. Functional Data Analysis (FDA) has emerged as an advanced statistical method for overcoming the shortcomings of traditional analytic methods, because the information about continuous signals can be assessed over time. This paper takes the current literature a step further by presenting a new time scale normalization method, based on the Hilbert transform, for the analysis of functional data and the assessment of the effect on the variability of human movement waveforms. Furthermore, a quantitative comparison of well-known methods for normalizing datasets of temporal biomechanical waveforms using functional data is carried out, including the linear normalization method and nonlinear registration methods of functional data. This is done using an exhaustive database of human neck flexion-extension movements, which encompasses 423 complete cycles of 31 healthy subjects measured in two trials of the experiment on different days. The results show the advantages of the novel method compared to existing techniques in terms of computational cost and the effectiveness of time-scale normalization on the phase differences of curves and on the amplitude of means, which are assessed by Root Mean Square (RMS) values of functional means of angles, angular velocities, and angular accelerations. Additionally, the confidence intervals are obtained through a bootstrapping process.

show abstract