SHWASA
CALIBRATING SIGNAL042%
LIVE RESEARCH PROTOTYPENOT FOR CLINICAL DIAGNOSIS

01 / Pipeline

From breath to signal.

A transparent audio workflow turns a short respiratory recording into two complementary research outputs.

CANONICAL SIGNAL JOURNEY

One recording. A chain of evidence.

Switch research tracks to see how the same respiratory sound becomes a different question and output.

Record audio16 kHz resampling
COPD-EFF · STAGE 1 / 7

Record audio

Start with the respiratory recording as inspectable evidence.

INPUT audio recordingOUTPUT Record audio
Binary COPD screeningResearch screening signal only. Not a clinical diagnosis.
LIVE ARCHITECTURESPACE TO PAUSE · ARROW KEYS TO STEP
THE MODEL, EXPLAINED

How sound becomes a signal.

Follow one recording through the system. Each stage changes the question the model is able to ask.

Chest sound / 1–10 sec
Time-domain waveform
01Capture
STAGE 01 / 04 · RAW SIGNAL

A breath becomes a measurable waveform.

The microphone records pressure changes through a short, repeatable chest-sound protocol. Nothing is classified yet: this is the unedited evidence.

INPUTChest sound / 1–10 sec
→
OUTPUTTime-domain waveform
WHY THIS MATTERS Traceability over black-box magic.Research prototype · not for diagnosis
Research artifactInteractive figure
Audio preprocessing workflowClick to inspect · Esc to close
Research artifactInteractive figure
Feature extraction workflowClick to inspect · Esc to close
01

Capture

Standardized chest sound recordings from a small, repeatable protocol.

02

Clean

Patient-aware splitting, resampling, normalization, and fixed windows protect the evidence.

03

Represent

Waveforms become model-specific time-frequency features.

04

Predict

Each track returns its own probability or sound-event class for review.

2 trackscomplementary outputs
16 kHzshared audio rate
5 secCOPD-EFF fixed clip
SHWASA · RESPIRATORY ACOUSTIC INTELLIGENCEOPEN METHODS / INSPECTABLE OUTPUTS