DCD kinematic signature

Description

Six primary outcomes were retained: peak based smoothness (S1), Spectral Arc Length (SAL), Log Dimensionless Jerk (LDLJ), normalized agility score, PeakSD, and normalized Shannon entropy. They were organized into smoothness, agility/regulation, and complexity domains.All six outcomes were calculated separately for each trial. Trial-level values were then averaged within participant and Criticality condition. The final analytical unit was Participant × Criticality, preventing repeated trials from being treated as statistically independent observations. The dominant-hand trajectory was digitized frame-by-frame from video recordings using Kinovea and processed in MATLAB R2023a. The camera was positioned perpendicular to the plane of movement, and the analysis was two-dimensional. Position data were inspected for digitization errors before filtering and numerical differentiation. A fourth-order zero-phase Butterworth low-pass filter with a 5-Hz cutoff frequency was applied before calculation of velocity, acceleration, and jerk. Forward and reverse filtering minimized phase distortion. The 5-Hz cutoff was retained as the prespecified processing condition. Filtering is particularly important for jerk-based outcomes because differentiation can amplify high-frequency noise (Balasubramanian et al., 2015). A 5-Hz cutoff was selected as it effectively preserves the fundamental frequency components of voluntary, goal-directed human upper-limb movements while attenuating high-frequency differentiation noise (Winter, 2009).

Steps to reproduce

Resultant hand velocity was calculated from the two-dimensional velocity components: v(t) = √(vₓ² + vᵧ²) Acceleration and jerk were subsequently obtained by numerical differentiation of the filtered signals. Peak-based smoothness (S1) Velocity peaks were detected from the absolute resultant velocity profile using a minimum temporal separation of 150 ms and a minimum peak height of 10% of the maximum absolute velocity, matching the implemented MATLAB procedure and the original peak-based framework. S₁ = (Npeaks,TD / Npeaks,i) × 100 Spectral Arc Length (SAL) The supplied MATLAB code calculates a discrete spectral-arc-length score. The discrete Fourier transform was: V[k] = Σₙ₌₀ᴺ⁻¹ v[n] e^(-j2πkn/N) The magnitude spectrum was normalized by its maximum: V̂[k] = |V[k]| / max|V[k]| The spectrum was truncated at the implemented 20-Hz upper limit. The discrete arc-length calculation was: Lspect = Σₖ₌₁ᴷ⁻¹ √((1/(K-1))² + (ΔV̂[k])²) The normalized score was: SALnorm = (Lspect,TD / Lspect,i) × 100 Log Dimensionless Jerk (LDLJ) LDLJ was calculated from the jerk signal after filtering and numerical differentiation. The exact implemented form was: DLJ = -√[(1/(T⁵Vpeak² + ε)) ∫ₜ₁ᵗ² jerk(t)² dt] LDLJ = -ln(|DLJ|) Normalized Agility Index The Agility Index was calculated as: A=γ(t)" " μ(Δv)" " P where Arepresents the Agility Index, Prepresents the precision component, and γ(t)and μ(Δv)are weighting functions related to movement duration and velocity characteristics, respectively. The velocity difference was calculated as: Δv=v_max-v_mean The temporal weighting component was calculated as: γ(t)=1+β_2 (1-t/t_ref ) where tis the movement duration, t_refis the reference movement duration, and β_2=0.2. The velocity-related weighting component was calculated according to the reference velocity-difference criterion described by de los Reyes-Guzmán et al. (2016), with β_3=0.05and Δv_ref=0.65: μ(v_"dif" )={■(1,&v_"dif" ≤v_"dif,ref" @1+β_3 (1-v_"dif" /v_"dif,ref" ),&v_"dif" >v_"dif,ref" )┤ The resulting Agility Index was normalized relative to the reference performance and expressed on a 100-point scale: A_norm=100 A/A_ref Acceleration Peak Standard Deviation (PeakSD) Acceleration peaks were detected from the absolute filtered acceleration profile using a minimum peak separation of 0.05 × Fs. PeakSD was calculated as: PeakSD = √[Σᵢ₌₁ⁿ(aᵢ - ā)²/(n-1)] ai is the magnitude of the ith acceleration peak, ā is the mean acceleration-peak magnitude, and n is the number of detected peaks. PeakSD represents variability in movement intensity/regulation rather than direct smoothness. Normalized Shannon entropy The velocity signal was discretized into 32 histogram bins. For bin b: p_b = n_b / N Shannon entropy was: H = -Σᵦ₌₁³² p_b log₂(p_b) Normalized entropy was: H_norm = H / log₂(32) Higher values indicate a broader distribution of velocity observations across the defined bins and were interpreted as greater distributional complexity.

Institutions

• University of Tehran

  Tehran, Tehran

Categories

Licence