Product information
Key Product Features
The following are some key features of the active voice biometrics product available through LumenVox containers:
Identity Management
The product allows the client to manage the identities against which a voiceprint model is stored. This includes the creation, removal, disabling and enabling functions along with the ability to retrieve the enrollment status of a given identity.
Voiceprint Enrollment and Verification
Through the APIs available the client can enroll or verify a voice from any channel (see list of possible channels above). provided they can record the voice and supply the audio in the required format.
Voice Biometrics - Deep Neural Network (DNN) Engine
Our voice biometrics engine is built using deep neural networks allowing LumenVox to create highly accurate models for active, passive and hybrid voice authentication use cases. This engine allows LumenVox to biometrically enroll and verify in any language using any passphrase (provided the language is catered for in the ASR engine for text validation). The engine can be used out-of-the-box or custom calibrated and tuned using the client’s voice biometrics data.
Automatic Speech Recognition (ASR) Engine
Our new speech recognition engine also makes use of deep neural networks allowing LumenVox to detect what is being said in each audio sample with extreme accuracy. This component is used to validate that the consumer is repeating the passphrase presented to them whereas the voice biometrics engine only determines if the same person is repeating a phrase.
Reporting
Through the reporting API or LumenVox web portal, the client can extract all session, transaction, audit, operation information as well as retrieve various statistics, counters, performance metrics & tuning indicators. The client can also view speaker enrolments and verifications, listen to audio, view metadata, and troubleshoot transactions.
Playback detection
The product features playback detection which detects if a recording presented to the engine is similar to one it has processed before. This is a an anti-spoofing measure used to detect if a potential fraudster is attempting to replay an audio sample of a speaker uttering the passphrase.
Liveness detection
This feature minimizes the risk of fixed phrases e.g., “my voice is my password” been recorded and played back. This fraud attempt is thwarted by the system processing random phrases uttered by the speaker. This is used in active use cases.
Synthetic speech detection
The product features identifies that the audio been recorded has been generated by some form of synthesizer. This is primarily used in passive use cases but could extend to active use cases.
Fraud detection
This feature allows for audio of a fraudster to be added to a watchlist, and any enrollments and verifications are checked against this alerting the client to possible fraud attempts.