GridID: Multimodal Biometric Identification System

GridID is a unified biometric identification and analysis platform designed for law enforcement agencies and forensic laboratories dealing with organized crime, fraud, corruption, terrorism and other types of security threats.
USE-CASE SCENARIO 1USE-CASE SCENARIO 2
Product Library:

Features

  • Streamlined process of biometric samples gathering and enrollment
  • Investigative functionality:
    • utomatic search and identification by voice, face and other biometric patterns
    • real-time multi-thread analysis and identification of voices incalls
    • voice data processing: keyword search in speech recordings, speech recognition (Speech-to-Text), speaker’s language and gender detection
  • Forensic analysis:
    • biometric data verification and reporting for legal proceedings
    • noise cancellation and speech enhancement
    • authenticity analysis of voice and video recordings
  • Report generation and monitoring performance of users and systems

Advantages

  • Full cycle of biometric analysis: from sample collection to a court report generation
  • Automatic language and gender identification, keyword spotting and speech-to-text conversion
  • Automatic speech detection and segmentation
  • Scalable centralized or distributed architecture
  • Ready for integration with external systems or additional biometric parameters
  • Automatic pre-processing and indexing of biometric patterns and quality control
  • Flexible modular structure to meet particular need of every departmen

Perfomance

Voice biometric

  • Up to 100k processed calls per day (for 16 cores processing server, avg. 2 min. calls duration)
  • Biometric processing time = 35RT . Up 10,000 biometric comparisons per sec (1 core)
  • EER of GridID.ScanRT = 5% (for 2 minutes average call duration)
  • EER of GridID.Scan = 1,5% (for comparison of models built on pure voice samples duration more than 96 sec)

Face biometric

  • Calculation of sample = 230 ms
  • Speed of comparison = 1000k per sec
  • Sample detection avg. error = 5% (of eye distance)
  • EER of verification = ~0.5%: FA (FR=10e-5) depending on quality of patterns