SciTransfer
SpeechXRays · Project

Voice-and-Face Identity Verification That Works on Standard Smartphones and Laptops

digitalPilotedTRL 7

Imagine unlocking your phone or logging into your bank not just with your face, but by speaking — and the system watches your lips move at the same time to make sure it's really you. SpeechXRays combined voice recognition with video analysis of lip movement and facial features to create a biometric security system that's twice as accurate as what's commercially available today. It runs on regular smartphone cameras and microphones, so no special hardware is needed. They tested it with 2,000 real users across three different scenarios: workplace access, healthcare, and consumer services.

By the numbers
1/100
Crossover accuracy — twice the accuracy of commercial voice/face solutions
2,000
Real users tested across pilot environments
3
Real-life pilot scenarios (workforce, eHealth, consumer)
EUR 4,102,467
EU contribution to the project
13
Partners in the consortium across 7 countries
35
Total project deliverables produced
The business problem

What needed solving

Passwords are insecure and inconvenient. Single-factor biometrics like voice-only or face-only can be spoofed with recordings, photos, or deepfakes. Current commercial biometric solutions deliver limited accuracy, require expensive specialized hardware, or store sensitive biometric data on central servers — creating privacy and compliance risks under GDPR.

The solution

What was built

Three successive system releases (v1, v2, v3) integrating multi-channel biometric recognition with user interfaces and privacy components. A context-dependent tuning framework that adjusts security thresholds based on the sensitivity of the service being accessed. The full system combines acoustic voice analysis with machine vision of lip movement and face recognition, running on standard embedded cameras and microphones.

Audience

Who needs this

Digital banks and fintech companies fighting identity fraud on mobile appsHospital systems and telemedicine platforms needing GDPR-compliant patient authenticationEnterprise IT departments replacing password-based or badge-based access controlGovernment agencies digitizing citizen services that require strong identity verificationSmart building operators looking for hardware-free biometric access solutions
Business applications

Who can put this to work

Banking & Financial Services
enterprise
Target: Digital banks and fintech companies offering mobile banking apps

If you are a digital bank dealing with rising identity fraud and customer friction during login — this project developed a multi-channel biometric system achieving 1/100 crossover accuracy (twice current commercial solutions) that runs on standard smartphone cameras and microphones. It was tested with 2,000 users across real-life scenarios, meaning your customers could authenticate with a glance and a spoken word instead of remembering passwords or waiting for SMS codes.

Healthcare & eHealth
enterprise
Target: Hospital chains and telemedicine platforms managing patient identity

If you are a healthcare provider struggling with patient identity verification for telemedicine consultations or electronic health records access — this project built and piloted an eHealth use case where patients authenticate using voice and face together. The system stores biometric data on the user's own device rather than central servers, directly addressing GDPR and health data privacy concerns. It was validated with 2,000 users across 3 pilot environments.

Corporate Security & Access Control
mid-size
Target: Large enterprises and facility management companies handling physical and digital access

If you are a facility management company or enterprise IT department dealing with badge cloning, shared passwords, or expensive access control hardware — this project developed a workforce authentication system using only embedded microphones and cameras already in laptops and phones. The context-dependent tuning framework lets you adjust security levels based on what the employee is trying to access, from opening a door to approving a financial transaction.

Frequently asked

Quick answers

What would it cost to deploy this biometric system?

The system was specifically designed for cost-efficiency by using standard embedded microphones and cameras found in smartphones and laptops — no special hardware required. Licensing and integration costs would need to be discussed with the coordinator IDEMIA, a world leader in digital security solutions. The project received EUR 4,102,467 in EU funding across 13 partners over 36 months.

Can this scale to millions of users in a production environment?

The system was piloted with 2,000 users across 3 real-life environments (workforce, eHealth, and consumer), proving it works outside the lab. Since it runs on standard device hardware and biometric data is stored locally on user devices (not centrally), the architecture avoids server-side bottleneck scaling issues. Moving from 2,000-user pilots to full commercial deployment would require further engineering by the consortium partners.

Who owns the technology and how can we license it?

The project was coordinated by IDEMIA France, a major digital security company (not an SME). With 9 industry partners including 5 SMEs across 7 countries, IP is likely shared among consortium members. Licensing discussions should start with IDEMIA as coordinator, given their existing commercial biometrics portfolio.

Does this meet GDPR and data protection requirements?

Privacy was a core design principle: biometric data is stored on the user's own device or in a private cloud under the responsibility of the data subject, not on central company servers. This privacy-by-design approach directly addresses GDPR requirements for biometric data processing. The eHealth pilot specifically tested this in a healthcare context with strict data protection rules.

How accurate is it compared to what we use today?

The project achieved a crossover accuracy of 1/100, which is stated to be twice the accuracy of commercial voice and face solutions available at the time. The system also provides superior anti-spoofing capabilities by combining voice acoustics with lip movement analysis — making it harder to fool with photos, recordings, or deepfakes.

How long would integration take with our existing systems?

The project produced three successive system releases (v1 through v3), each integrating biometric components with user interfaces and security/privacy modules. A context-dependent tuning framework was also delivered that allows matching thresholds to be adapted based on the security level required. Based on available project data, integration timelines would depend on your existing identity management infrastructure.

Consortium

Who built it

The consortium is heavily industry-driven with 9 out of 13 partners (69%) coming from the private sector, including 5 SMEs, backed by 2 universities and 2 research organizations. Coordinated by IDEMIA France — one of the world's largest digital identity and security companies — the project had serious commercial intent from the start. The geographic spread across 7 countries (DE, EE, EL, FR, HU, RO, UK) gives broad European market access. With EUR 4,102,467 in EU funding and an Innovation Action classification, this was designed to move technology toward market, not just publish papers.

How to reach the team

IDEMIA France — global digital security company, formerly known as OT-Morpho/Safran Identity & Security. Reach the project lead through IDEMIA's biometric solutions division.

Next steps

Talk to the team behind this work.

Want to explore licensing this multi-channel biometric technology for your organization? SciTransfer can connect you with the right people at IDEMIA and the consortium partners. Contact us for a warm introduction.