Inovix Inc.

Inovix VoiceVerify SDK

In everyday life the mobile phone is about to become one of the most important tools. It certainly has spawned an astonishing range of applications. As the mobile phone becomes connected to more aspects of personnel and financial life as well as to the enterprise structure the risks of mobile phone fraud are growing.

The Inovix VoiceVerify SDK is designed for OEMs, VARs and developers who want to easily integrate voice matching technology into both existing and new applications. This product will enable businesses to add security to applications in a wide range of areas, including data-security, telecommunications, financial services, e-commerce and healthcare services.

Available for 2 platforms:

• Inovix Fingerprint SDK for PC.
• Inovix Fingerprint SDK for Win CE handhelds.
• coming soon: Android and iPhone releases.
  

Features:

• Device independence: Inovix VoiceVerify SDK accepts as input a 16 bits per sample audio buffer in uncompressed format (.wav pcm encoding) recorded at a frequency of 8 KHz or more. The sound buffer can come from any recording device. This allows the end-user or system integrator to choose the most suitable device for their application. Inovix recommends using a good quality microphone for best results.

• The SDK is built using a multilayer technology that includes: noise canceling, feature extraction, channel normalization, model building, score normalization across different handsets and decision fusion logic.

•  Inovix VoiceVerify SDK’s world class algorithm ensures speed, accuracy and reliability of the voice template extraction and matching. The handheld and embedded version of the algorithm uses only fixed point arithmetic operations to deliver real-time speed.

•  With Inovix VoiceVerify SDK voice password resets and rotation policies at the biometric level can be successfully employed.

Inovix VoiceVerify SDK encompasses a complete suite: Text Independent (TI), Text Prompted (TP) and Text Dependent (TD) speaker recognition engines.

TI engine can operate using a variety of information such as account numbers, names, secret passwords and random phrases uttered in any language. Users can be enrolled and verified on the same information or they can use different information for enrollment and verification.

TP engine prompts the user during the enrollment session to read a few preselected digit sequences (such as 01234). Training session lasts under 1 minute. Only English language can be used. At verification the user is prompted for a random digit sequence. Only the prompted digit sequence can be used to verify the user.

For the TD engine a user is required to enroll by pronouncing a pass-phrase just a few times (such as "21-67-89" or First+Last Name). Key phrase duration is no longer than 2 seconds. Training session lasts less than 30 seconds. Pass phrase is language independent. TD technology accepts only the pass-phrase used at enrollment for verification.

These capabilities offer the developer total control over the enroll and verification prompts. The user is guided trough a simple enroll process at the end of which a speaker template is created and stored. At verification time the user's voice is compared against the stored speaker template. The enrollment and verification operations can be run on a PC, PDA or Phone device.


How to use (a simple 2 step process):

1. Use the Enroll class to voice enroll a user and create the speaker template.
2. Use the Verify class to voice verify a user and make the accept/reject decision.

Specifications:

Inovix Fingerprint SDK for Windows PC

Inovix Fingerprint SDK for Windows CE handhelds:


To commercialize this technology we are looking for investors & system integrators which can be solution and design partners and can help with commercial application marketing, sales and deployment.

See our "Security in Mobile Markets" white paper for potential applications.

Demo sample showing the Inovix Fingerprint SDK capabilities is available as a free download. Coming soon!