BioVox is our speaker verification and identification product. Thanks to its advanced voice biometrics technology you can raise the security level in physical access controls or logins and at the same time get rid of magnetic cards and passwords. So you have a double gain: increased security + user friendly systems.
A critical feature of BioVox that stands out of other voice biometrics technologies is its ability to be text independent. Thanks to this, a new user can enroll into the system saying whatever text he or she wants, or by using an existing recording, and later get identified speaking a totally different sentence, even in another language! This feature almost completely terminates the risk of spoofing attacks, because the user can be asked to say a random text captcha-style. In addition, it allows a much wider range of applications beyond access controls, like searching for speakers in large databases of audio recordings or continuous identity verification of the speaker along all the duration of a phone conversation, just to name a couple of possibilities.
BioVox is not a closed architecture application with a predefined graphical user interface but an open development platform that exports its functionalities through a powerful yet easy to use API (Application Programming Interface). With this API you can integrate a complete voice based user validation system into any embedded hardware or software application.
BioVox has been fully developed by DTec.
The API provides all the functionality that is needed for the integration and maintenance of a fully functional voice based user validation system, from the initial stage of user enrollment, where the voice signature is created, to the final validation step, where the actual user verification takes place or the closest matching signature is returned, if an identification task is performed.
Besides the most usual family of security applications aimed to physical or logical control, it’s also possible to identify an individual that interacts with our system in order to automatically adapt to his or her preferences, or to dynamically create custom content. In this case, although an initial enrollment step is always needed, the user doesn’t provide his or her supposed identity to be verified by the system. Instead of that, the user speaks directly pronouncing a single word or a full sentence and then the system automatically performs a matching process against all the voice signatures available choosing the best match, if any.
A BioVox based system can be naturally coupled with our speech recognition product: ReconVox. This way you can implement an IVR (Interactive Voice Response) service capable of continuous speech understanding, enhanced with automatic speaker verification working throughout the call, in the background and in a transparent fashion, for example. Thus security is exponentially increased in every point of the transaction. Or you could develop electronic devices reacting to voice commands just issued by the authorized user.
- Double role: verification (physical or logical access control) and identification (suspect searching or automatic customization).
- Text and language independent.
- Security level can be adjusted.
- Automatic measurement of the quality of the voice signature just created.
- Two operation modes: live or batch mode.
- Highly optimized verification engine: can be integrated into embedded systems.
- Available both in DLL (Dynamic Link Library) format for Windows and in shared object format for UNIX/Linux. Please ask for other platforms.
- Security in electronic and phone transactions (e-banking, user license validation, calls to Public Administration services…).
- e-commerce (secure payment in Internet, payment with mobile phones…).
- Physical access controls and presence controls (no more buddy punching).
- Telepresence control (GPS location + voice biometrics, secured remote voice commands…).
- Alarms and domotics (electronic devices driven by the voice of authorized users, alarms (de)activation…).
- Police investigations, forensics and intelligence (virtual identification parades, suspect identification in audio recordings…).
- Electronic devices driven by voice in help for the impaired people (voice commands just coming from the authorized user).
- Security for automobiles (engine start with the voice of the owner, navigator and phone calls controlled by voice…).
- Audio/video search by content (search for specific speakers in audio recordings or live broadcasts).
The term biometrics is applied to the set of techniques used to automatically identify individuals through the analysis of behavior or physical features. This recognition is performed based on different features of the human body that have unique values for every person: fingerprints, hand shape, face geometry, voice, retina, iris…
Every biometric system works in two successive steps: the enrollment step, when the physical features of the individual are extracted and used to calculate a biometric certificate or signature, and the validation step, when the actual verification of the user takes place, analyzing his or her face geometry, fingerprint, voice, etc… and performing a matching process against his or her biometric signature.
There are two main families of biometric applications, taking into consideration the specific needs of the validation step:
- Verification: the task is aimed to check whether the speaker is actually whoever he or she says he or she is. Therefore, the output of the biometric system is the validation or rejection of the individual (one to one matching).
Identification: this time the task is aimed to decide whether the individual, whose identity is unknown a priori, is included in the repository with all the biometric signatures generated in the enrollment step and, if this answer is affirmative, which one it is (one to many matching).
Different biometric techniques fit better into different types of applications. For example, voice biometrics fits perfectly whenever mobility is the key, like in phone banking, e-commerce and in any mobile applications, because all the hardware needed is a conventional microphone like the ones in mobiles or in personal computers.
However, iris recognition requires special and expensive hardware in order to capture the high quality images needed. In addition, they usually have to be captured in favorable environment conditions. Therefore it’s best reserved for very high security physical access controls.