This pilot study investigates the feasibility of objectively quantifying movement amplitude and amplitude decay during large-amplitude exercises using RGB camera-based video analysis in healthy volunteers. Six large-amplitude exercises are recorded across repeated trials, with the first repetition clinically confirmed as the individual's maximum reference. Movement amplitude is quantified using pose-based kinematic measures and complemented by an exploratory self-supervised visual transformer approach to characterize repetition-wise changes relative to the reference execution. The study aims to determine whether camera-based methods can detect amplitude decay patterns and identify moments requiring clinical verbal cues, thereby supporting future objective monitoring and feedback in large-amplitude exercise training.

Parkinson's disease is characterized by bradykinesia and impaired motor scaling, leading to reduced movement amplitude and diminished functional performance. Large-amplitude exercise approaches are widely used to counteract these deficits by promoting exaggerated movement execution and recalibration of internal movement perception. However, during repetitive exercise sets, individuals often demonstrate a gradual reduction in movement amplitude, a phenomenon known as amplitude decay, which typically requires verbal correction by a clinician.

Despite its clinical relevance, amplitude decay has not been objectively quantified, and movement quality is commonly assessed through visual observation alone. Marker-based motion capture systems, while accurate, are impractical for routine or remote use. RGB camera-based markerless analysis offers a low-cost and accessible alternative for objective movement assessment, particularly in telerehabilitation contexts where direct visual supervision is limited.

Participants and Study Design

A total of 22-26 healthy volunteers aged 18-45 years will be recruited. All participants will be capable of independent functional movement and will provide written informed consent prior to participation. The study follows a single-session, observational design.

Exercise Protocol

Participants will perform six standardized large-amplitude exercises:

Floor-to-Ceiling Reach

Forward Step

Side Step

Sit-to-Stand

Forward Reach to Grasp and Release an Object

Big Walking

Each exercise will be recorded for eight repetitions using a single RGB camera capturing full-body movement. No wearable sensors or reflective markers will be used.

For each exercise, the first repetition will be performed without prior instruction regarding maximal effort. Immediately after completion, a clinician expert will verbally confirm whether this execution represents the participant's maximum movement amplitude. Upon confirmation, this repetition will be designated as the individual reference trial for that exercise.

Data Processing and Outcome Measures

Movement data will be analyzed using two complementary approaches:

Pose-Based Analysis:

Joint keypoints will be extracted from video using a pose estimation algorithm. Movement amplitude will be computed using joint displacement-based metrics and normalized according to individual anthropometric characteristics. Amplitude decay will be quantified by comparing early and late repetitions and by calculating repetition-wise trends.

Visual Representation Analysis:

As an exploratory component, a self-supervised vision transformer will be used to extract visual feature representations from exercise videos. Each repetition will be compared to the clinician-confirmed reference repetition using feature similarity measures. Repetition-wise similarity scores will be used to characterize within-exercise changes and identify patterns consistent with amplitude decay.

Statistical Analysis

Data distribution will be assessed for normality. Changes in movement amplitude and repetition-wise trends will be analyzed using paired statistical methods. Associations between camera-based measures and clinician observations will be explored using correlation analysis, with statistical significance set at p < 0.05.

Ethical Considerations and Expected Impact

The study involves no invasive procedures or therapeutic interventions. Video data will be anonymized and securely stored, and participants may withdraw at any time. Ethical approval has been obtained from the Biruni University Non-Interventional Ethics Committee.

This study is expected to demonstrate the feasibility of camera-based methods for objective assessment of movement amplitude and amplitude decay, providing a methodological foundation for future studies involving patient populations, automated feedback systems, and telerehabilitation applications

Inclusion Criteria:

Ability to stand, walk, reach, step, and perform sit-to-stand movements independently without assistance.

No self-reported or clinically diagnosed neurological, musculoskeletal, or vestibular disorders.

Normal or corrected-to-normal vision sufficient to follow visual instructions.

Exclusion Criteria:

• History of neurological disorders (e.g., Parkinson's disease, stroke, multiple sclerosis, traumatic brain injury).

History of major musculoskeletal injury or surgery affecting upper or lower extremities that limits movement.

Presence of balance disorders or vestibular dysfunction.

Cardiovascular, respiratory, or metabolic conditions that may limit safe participation in physical exercise.

Current pain, dizziness, or excessive fatigue during movement or exercise.

Pregnancy or any condition for which moderate-intensity physical activity is contraindicated.

Use of assistive devices for walking or balance.

Refusal or inability to provide consent for RGB video recording.

This study employs single-level masking at the participant level. Participants are blinded to the specific analytical objectives and outcome measures of the study. While participants are informed that their movements are being video recorded during large-amplitude exercises, they are not informed about the criteria used to evaluate movement amplitude, repetition-wise changes, or the reference-based analysis approach.