Multimodal emotion recognition · PHQ-9

Reading well-being from a face, a voice, and a questionnaire.

A research demo that pairs deep-learning emotion recognition with the clinically validated PHQ-9 — all running in your browser, nothing stored.

01

Facial emotion

Your webcam streams to a convolutional network trained on FER2013, classifying expression frame by frame.

Open →
02

Speech emotion

Record a few seconds; a 1-D CNN trained on RAVDESS predicts a combined gender and emotion class from your voice.

Open →
03

PHQ-9 screening

The standard nine-item depression-screening questionnaire, scored with severity bands and guidance.

Open →

What this is

A demo of multimodal well-being screening

What it's for

A hands-on demonstration of how machine learning can read emotional cues, paired with a validated questionnaire. Useful for:

  • Quick self-reflection & mood check-ins — gauge how you're presenting right now.
  • Showcasing an end-to-end ML pipeline — vision + audio models behind a web app.
  • Education & awareness — see what affect-recognition can (and can't) do.

How to use it

  1. Pick a modalityFace (webcam), Voice (mic), or the PHQ-9 form.
  2. Run it in your browserGrant camera/mic access — nothing is uploaded or stored.
  3. Read the resultAn emotion cue, or a scored PHQ-9 severity band with guidance.

Emotion recognition is a supplementary signal, not a diagnosis. The PHQ-9 is the only clinically validated component, and it is a screening aid — not a substitute for a clinician.