AI Walking Analysis Instructions

Record a steady habitual walk, choose the correct view and assessment, then review overlays, metric cards, and a shareable PDF summary.

The walking app supports Side View and Front / Back analysis plus three clinical modes: Parkinson's Disease gait assessment, Fall Risk & Frailty screening, and Orthopedic Recovery. Inside the app you can upload a video, enter the patient height, review pose overlays and color-coded result cards, and share or download a results report.

Side View Front / Back PD Gait Assessment Fall Risk & Frailty Orthopedic Recovery PDF Results

Open Walking App Back to Homepage

How to record your walking video for gait analysis

Recording Setup

Capture a consistent walking clip before you upload.

The most reliable analysis comes from a short clip that shows several uninterrupted steps at the person's natural pace. The first figure summarizes the most important filming rules and the app depends on those details more than most people expect.

Warm up first: walk for about 3-5 minutes so the gait settles into a natural, habitual pace. Keep the pace consistent during the recorded clip.
Use a stationary camera: film from a tripod or stable support at about hip height. The walk figure recommends HD 1080p and 60 fps or higher when possible.
Keep the full body visible: head, trunk, hips, knees, ankles, and feet should stay in frame for multiple consecutive steps on a straight path.
Avoid distortion: use a normal lens or modest zoom, roughly 1x to 3x. Avoid fisheye or ultra-wide views because they can distort stride and posture.
Make the joints easy to see: use bright lighting and clothing that does not hide the shoulders, hips, knees, ankles, or wrists.
Choose the correct view: record Side View when you want forward progression, timing, trunk posture, and knee-motion metrics. Record Front / Back when you want symmetry, arm swing, step width, and trunk sway metrics.
Redo poor clips: if the feet leave frame, the subject turns, another person blocks the view, or the pace changes, record again before interpreting the results.

Important: enter the patient height before judging distance-based values. The walking app uses height to estimate body scale, pixels per meter, walking speed, stride length, and some front/back distance notes.

Inside the App

Choose the view, select the assessment type, upload the walking video, review the overlays and metric cards, then use the Results menu to share or download a PDF summary.

Best Clip Checklist

Use several consecutive steps, a straight path, a steady habitual pace, and a subject who remains the most prominent person in frame from start to finish.

Start Walking Analysis

Side View

Use the side view for forward progression, timing, posture, and sagittal mechanics.

The side-view summary combines general walking mechanics from the figure with the assessment-specific metrics used in the app. This is where the platform measures stride, speed, double-support timing, trunk posture, and knee motion.

Core Rhythm

Cadence, stride length, and double-support time describe how efficiently the gait cycle turns over and how long each step spends in overlap.

Impact & Efficiency

Walking speed, foot strike velocity, and vertical oscillation help show whether the person is moving forward smoothly instead of wasting motion.

Sagittal Kinematics

Trunk lean, hip extension, arm swing amplitude, and mid-swing knee flexion help explain posture, stiffness, and smooth roll-through.

Joint Loads

The figure also shows knee and ankle loading estimates, which can help frame joint stress during walking even when the main clinical interpretation centers on other metrics.

Parkinson's Disease gait assessment: absolute stride length is generally Good > 1.1 m, Okay 0.8-1.1 m, and Poor < 0.8 m. Trunk anterior flexion is generally Good 0-10 deg, Okay 11-20 deg, and Poor > 20 deg.
Fall Risk & Frailty screening: habitual gait speed is generally Good > 1.0 m/s, Okay 0.8-1.0 m/s, and Poor <= 0.8 m/s. Double-support time is generally Good < 25% of the gait cycle, Okay 25-30%, and Poor > 30%.
Orthopedic Recovery: stance phase duration asymmetry is generally Good < 3% difference, Okay 3-8%, and Poor > 8%. Mid-stance knee flexion asymmetry is generally Good < 5 deg side-to-side difference, Okay 5-10 deg, and Poor > 10 deg.
Overlay guidance: the app can mark same-foot contacts, show a horizontal travel ruler with a live speed display, highlight double-support overlap, show the live trunk segment against vertical, and compare knee angles side to side.
How the cards read: results are color-coded Good, Okay, or Poor and can include the current value, target guidance, typical range, why the metric matters, and short improve or further-assess suggestions.

Open Side-View Analysis

Posterior-view walking analysis for gait and balance metrics

Front / Posterior View

Use the front or back view for symmetry, balance, and compensatory control.

The posterior-view summary focuses on the balance-oriented measures that are easiest to miss from the side. In the app, these results are especially useful for asymmetry screening, lateral instability, and compensatory unloading patterns.

PD Marker

Arm swing amplitude asymmetry can flag unilateral rigidity. The app traces left and right wrist travel and reports the side-to-side mismatch.

Fall Risk Marker

Step width variability captures how stable or erratic foot placement is from step to step during walking and is especially useful for lateral balance screening.

Orthopedic Marker

Compensatory trunk sway reflects lateral torso lean during stance and is often interpreted alongside stance asymmetry rather than in isolation.

Detection Detail

The figure notes that the system uses COCO 17 keypoints and shows three headline metrics, but the app also layers pose overlays and color-coded assessment cards onto the selected video.

Parkinson's Disease gait assessment: arm swing amplitude asymmetry is generally Good < 10% side-to-side difference, Okay 10-20%, and Poor > 20%.
Fall Risk & Frailty screening: step width variability is generally Good < 4% coefficient of variation, Okay 4-8%, and Poor > 8%. High variability suggests unstable lateral control and higher fall risk.
Orthopedic Recovery: compensatory trunk sway is generally Good < 5 deg lean, Okay 5-10 deg, and Poor > 10 deg. Excessive sway can reflect hip abductor weakness, pain, or joint unloading.
Overlay guidance: the app can trace wrist travel for arm swing, sample ankle separation at step contacts for variability, and draw the torso against vertical to report stance-phase sway.
Useful follow-up cues: when asymmetry or sway stays high, the app's coaching language points toward reciprocal arm swing drills, hip abductor and trunk strengthening, and level-pelvis or upright-trunk cueing during stance.
Distance note: front and posterior distance-based interpretations still depend on estimated body scale from visible segments, so clear framing and correct height entry matter here too.

Open Front / Back Analysis

Clinical Use, Results Workflow, and Safety

Use this as a screening and follow-up tool: walking analysis from video can be extremely helpful, but it should be interpreted together with symptoms, clinical examination, strength or balance testing, and patient history. It is not intended to function as a stand-alone diagnosis.

Results workflow: after analysis, the app can generate a styled PDF summary that includes the selected assessment, the current view, the same color-coded feedback cards shown in the app, and a link back to the walking instructions. On devices without native sharing, the share action falls back to a download.

Recording quality still drives accuracy: clipped feet, hidden joints, curved walking paths, changing pace, poor lighting, or incorrect height entry can distort speed, stride, symmetry, and sway values. If the video quality is questionable, record again before making a clinical judgment.

Launch Walking App Back to Homepage