The International Institute for Integrated Performance

SOUND IN. PREDICTION OUT.

We translate the acoustics of a live crowd into prediction — a signal you can feel before the roar. Making live events accessible to everyone.

5-stagesignal pipeline
Classicalsignal processing
Openacoustic standard
Activereal-match validation

Hear It Coming. Before It Happens.

This is the actual AEGIS decision loop running in your browser: rolling crowd energy, a venue-adaptive σ threshold, and a detection trigger. Press Simulate goal moment and watch AEGIS call it before the peak.

AEGIS ENGINE — LIVE
Crowd energy
0.18
σ threshold
Swells detected
0
Lead time

Classical acoustic detection — STFT features, venue-adaptive sigma thresholding, RF classification. No wearables required for this demo; in the field, this trigger drives a haptic pattern on the wrist.

Real Technology. Being Validated in the Open.

The detection pipeline above is real — a working adaptive-threshold detector and a trained classifier, not a mockup. What we won't do is publish a precise accuracy or lead-time figure until it's been checked against real, independently-verified match data rather than the pipeline scoring itself. That re-validation is actively underway now. Specific numbers will be published here once they're real and defensible — not before.

Five Stages. One Signal.

Stage 01 · Input

Audio Feed

Continuous venue audio — broadcast feed or ambient capture. No cameras, no tracking, no personal data. Sound only.

SAMPLE RATE 44.1kHz · MONO SUFFICIENT
Stage 02 · Features

STFT + Cepstral Smoothing

Short-time Fourier transform decomposes the roar into frequency bands; cepstral smoothing strips noise while preserving the shape of the swell.

CROWD BAND ISOLATION · PA / WHISTLE REJECTION
Stage 03 · Detection

Venue-Adaptive σ Thresholding

Every stadium is loud differently. The threshold learns each venue's baseline and flags energy that departs from it — the collective inhale before the moment.

ROLLING μ + k·σ · SELF-CALIBRATING PER VENUE
Stage 04 · Classification

Random Forest Classifier

Is this swell a goal build-up, a near-miss, or noise? The RF classifier decides in milliseconds, trained on real labeled crowd-audio events. We're re-validating accuracy against independent match data before publishing a number.

CLASSICAL ML · QAOA VALIDATION AS SEPARATE OPTIONAL LAYER
Stage 05 · Output

Haptic Signal

The prediction becomes touch: rise, build, peak on the wrist — ahead of the stadium erupting. A blind fan feels the moment gathering with everyone else.

5-STATE VOCABULARY · PULSE & RALLY (CAD-STAGE)

Open Standards. Real Research.

ASI Acoustic Packet Standard

PUBLISHED APACHE-2.0 · 06·23·2026 · SEP CLAIMS C1–C6 IN PREPARATION

Written to Survive Peer Review.

WP·01

Acoustic Swell Detection in Live Sports Environments

Methodology overview.

In preparation
WP·02

AEGIS Validation: Real-Match Detection Accuracy

Independent validation against real, sourced match data — in progress.

In preparation
WP·03

Venue-Adaptive Sigma Thresholding for Crowd Noise

Technical methodology paper.

In preparation
PROGRAM · ACCESSIBILITY

VisionLink Pilot

IRB-ready protocol, co-designed with blind and low-vision sports fans in Philadelphia. Partnered with the National Disability Institute. Two beta testers confirmed and counting.

Pilot program →
PROGRAM · CONSULTING

Venue Accessibility Consulting

Audits, assessments, and advisory retainers that make stadiums genuinely accessible — not just compliant. Built on the same research that built AEGIS.

Consulting services →
EDUCATION

FIFA Agent Prep

Independent exam preparation for the FIFA Football Agent exam — 8 modules, full mock exam. Not affiliated with FIFA. $997 self-paced / $1,497 live cohort.

See the course →