MOVO-X
Loading MOVO-X…
Returning patients walk up to the kiosk, look at the camera, and are recognised in under 2 seconds — no card, no phone, no typing. Face recognition check-in is the fastest possible patient identification for high-volume returning-patient clinics.
Repeat patients at high-frequency clinics (dialysis, physiotherapy, oncology) visit dozens of times per year. Making them present their IC card every visit is unnecessarily slow and invasive once their identity is established. Face recognition check-in converts every subsequent visit into a 15-second arrival.
Patient face captured by kiosk camera. Matched against the encrypted face template stored from their first visit. Match completes in under 500 ms on the kiosk hardware — no cloud round-trip.
Anti-spoofing check rejects photos, phone screens, and 3D masks. Passive liveness — no blink or head-turn required. Works with glasses, headscarf, surgical mask partially removed, and low-light waiting areas.
On the patient's first visit, the kiosk captures a face template after NFC ID verification. Subsequent visits use face-only recognition. Enrolment takes 5 seconds and is transparent to the patient.
Match confidence threshold configurable per clinic. Low-security settings accept matches above 85%; high-security (controlled substance clinics) require 97%+ before proceeding without secondary verification.
Face templates are encrypted, stored in the clinic's own database, and never shared with third parties. Template deletion is one-click from the patient record or via automated PDPA deletion request.
If face match fails below threshold (poor lighting, head covering, camera occlusion), the kiosk falls back silently to NFC or manual check-in. No patient is blocked or embarrassed.
15-second check-in for returning patients using face recognition
Zero IC card presentation required after first visit
Liveness detection blocks 100% of static-photo spoof attempts in validation runs
Opt-out rate below 5% in clinics that explain the privacy-by-default model clearly
Face recognition uses MobileFaceNet architecture fine-tuned for Southeast Asian demographics and deployed as a TensorFlow Lite model on the kiosk NPU. Face templates are 512-dimensional embeddings — not raw images. Embeddings are stored encrypted (AES-256) in the clinic's Supabase instance. The kiosk computes a cosine similarity between the live embedding and the stored template; the result is a score between 0 and 1 compared against the threshold.
Yes. Face recognition check-in requires explicit opt-in consent on the first visit. Patients who do not consent use NFC or manual check-in on every visit — the service is never degraded for non-consenting patients.
No. The kiosk captures the face, computes the embedding, and discards the raw image immediately. Only the 512-dimensional embedding is stored — it cannot be used to reconstruct a face image.
At the default 92% threshold, false acceptance rate is below 0.01% in validation runs. This means approximately 1 in 10,000 match attempts could incorrectly identify a patient — comparable to existing manual ID-check error rates.
Yes, with the high-confidence threshold (97%+) configuration. Below threshold, the system requires NFC ID verification before proceeding — suitable for controlled substance dispensing workflows.
Template deletion is available via the patient portal, the clinic admin dashboard, or an automated PDPA deletion request flow. Deletion is immediate and irreversible — the patient returns to NFC/manual check-in from that point.
Tell us about your operation. We'll send a tailored proposal — architecture, integration scope, deployment timeline, and total investment — within hours.