VMM-1™ Voice Biometric Engine

Voice Biometric Group's state-of-the-art VMM-1™ voice biometric decision engine is the result of years of research and practical, field-based experience. Our next generation VMM-1™ engine uses statistical pattern matching techniques, advanced classification methods, and inputs from multiple mathematical algorithms to properly verify or identify speakers. The basic processes performed by the VMM-1™ engine are:

Analyis and Filtering.  All audio samples are thoroughly evaluated before they are sent to the voice biometric decision engine. This ensures that the requested enrollment, verification, or identification process will be successful. Different signal filtering techniques can be invoked as appropriate.

Feature Extraction.  The unique audio characteristics of speech are extracted as vocal features, modeled, and synthesized into voice prints. We use a combination of MFCC, LPC, and our own proprietary feature set within the VMM-1™ engine.

Algorithms.  To improve accuracy and reliability throughout real-world operating conditions, the VMM-1™ engine uses a number of different decision algorithms: DTW, LBG, GMM, and SVM. We work with our clients to configure the optimal mix of algorithms based on their unique needs.

Summary of Features: