Biometric Video
Intelligence
Identify. Track. Reconstruct.
Meet Kinerva
The open-source platform for multi-modal biometric video intelligence. Combining face, body, and gait recognition for high-confidence identification, tracking and kinetic analysis. Real-time and post-incident intelligence for security operations and investigations.
- Works when facial ID fails
- 160+ extracted features
- Court admissible evidence
Who is Kinerva built for?
Forensic / Law Enforcement
- Identify individuals from low-quality, distant, or night-time video
- Match suspects when faces are hidden or obscured
- Produce explainable, defensible results for court use
- Build suspect shortlists and connect crime scenes from large video datasets
Security / Access Control
- Re-ID individuals across multiple cameras and locations
- Maintain tracking when faces are not visible
- Add a second biometric layer to existing access control systems
- Detect and follow persons of interest in real time
Biomechanics / Medical
- Extract high-dimensional movement and body data from video
- Analyze gait, posture, and motion patterns at scale
- Enable research, diagnostics, and performance optimization
- Replace expensive motion capture setups with video-based analysis
How Gait & Body Recognition Works
3D Body Modelling
#1With our proprietary AI models we locate the exact positions of joints and create very accurate 3D digital twins of people appearing on standard CCTV footage.
Gait & Body Analysis
#2Personalized biomechanical models are applied to the 3D digital twins to extract movement patterns. Static and dynamic anthropometric features together are used to build personal biometric models.
Biometric Identification
#3Our system compares anthropometric features extracted from archived footage with those detected in live video or crime scene recordings, calculating a biometric match probability score based on over 100 objective metrics.
Turn Any Footage Into Evidence
When facial recognition fails, Kinerva identifies individuals based on their kinetic signature. Extracting gait and body characteristics from standard CCTV, it delivers court-ready, explainable evidence from crime scene footage.
- Identify suspects from low-quality, distant, or night-time video
- Match individuals even when faces are hidden or obscured
- Produce explainable, defensible results for court use
- Build suspect shortlists and connect crime scenes from large video datasets
Why we are the leaders in Motion Analysis
Our Approach to Movement Analysis
Typical Approach of Others
Device-specific sensor data preprocessing to reconstruct original movement
Sensor Data PreProcessing
Standard, general purpose data smoothing methods
Delicate balance of bespoke features artfully emulating key elements of human motor programs
Feature Space
Standard physical features, or observable, domain-specific features
Proprietary AI toolchain designed for motion analysis
Machine Learning
Popular, general purpose Python AI libraries or statistics modules
Our Approach to Movement Analysis
Typical Approach of Others
Possibility to handpick features when only limited amount of data is available
Data Size
Deep Learning on large amount of data only
Explainable models built on meaningful features
Prediction Models
Unexplainable “Black Box” models
Cross-domain knowledge</strong> from having analyzed:
- handwriting
- cursor movement
- video-based fine and gross motor movements
- other time-series data
Millions of users served by our solutions across:
- user identification
- signature verification
- personality profiling
- assessing neurological conditions
Sensor Data PreProcessing
Our Approach to Movement Analysis
Device-specific sensor data preprocessing to reconstruct original movement
Typical Approach of Others
Standard, general purpose data smoothing methods
Feature Space
Our Approach to Movement Analysis
Device-specific sensor data preprocessing to reconstruct original movement
Typical Approach of Others
Standard, general purpose data smoothing methods
Machine Learning
Our Approach to Movement Analysis
Device-specific sensor data preprocessing to reconstruct original movement
Typical Approach of Others
Standard, general purpose data smoothing methods
Data Size
Our Approach to Movement Analysis
Possibility to handpick features when only limited amount of data is available
Typical Approach of Others
Deep Learning on large amount of data only
Prediction Models
Our Approach to Movement Analysis
Possibility to handpick features when only limited amount of data is available
Typical Approach of Others
Deep Learning on large amount of data only
Cross-domain knowledge</strong> from having analyzed:
- handwriting
- cursor movement
- video-based fine and gross motor movements
- other time-series data
Millions of users served by our solutions across:
- user identification
- signature verification
- personality profiling
- assessing neurological conditions
Who we are
We at Cursor Insight are experts in AI-backed human movement analysis. Having invested a combined 100+ person-years into developing AI prediction models, we are leaders in our field.
Our award-winning machine learning technology is capable of identifying and classifying users by learning their unique movement patterns while they interact with a computer or a phone or appear on video. Based in the UK and Hungary, we build biometric applications across banking, forensics, cybersecurity and healthcare, serving millions of users.
Remote Monitoring Security Solution of the Year
Our partners
Featured in
Beyond the Limits of
Facial Recognition
Works when:
- faces are masked or obscured
- lighting is poor
- video quality is low
99.9%+
Accuracy
- 120+ static parameters
- 40+ dynamic features
- Reliability on par with facial recognition
Admissible
Evidence
- Identified a suspect in a real murder case in the EU
- Validated, explainable AI models
- Match scores based on up to 100+ objective metrics