{VSS Logo}
"Keypad Demonstration with Voice Protect® Technology"
VSS SDK Version 3.2 | Custom Hardware Prototypes | Smart Card Samples

 

Voice Protect® Keypad Security Simulator

This Active-X control demonstrates a simple speaker verification process without transmitting anything over the Internet or requiring a complicated series of installations and rebooting to try it. The entire download is less than 200K. All of your biometric information will stay in memory on your PC until the keypad is closed, at which time all of your biometric templates, your "voice prints", will be discarded.

This application, which includes a scaled down implementation of our patented voice authentication method, as well as the ability to run automated testing against over 450,000 voices in our test database is included. It has been written entirely using Microsoft Visual Basic 5.0.

System Requirements:

* PC running Windows 95/98/ME/NT/2000/XP.

* Microsoft Internet Explorer 4.01 or higher (Netscape does not support Active-X)

* Windows compatible microphone for speech input (calibrated).

The keypad format of the demonstration is an example of how the most commonly used method of human authentication used in the world, PINs, can be easily enhanced to include the use of a biometric "password". The "password" is provided by the person at the moment of a secure transaction or when accessing a secured area. In this case, a voice print.

This demonstration is simulation a real-world use of this technology in a residential or commercial keyless access device. This type device can be used to provide secure access for use in garage door openers, automobiles, cell phones, door locks, point of sale transactions, etc.

The method of speaker verification used in this sample is a re-creation of our original TFA analysis combined with our proprietary scoring and feature classification techniques, for which we hold two patents with a third pending. The algorithm used to encode template data is capable of storing the unique features of a single 2.8 second phrase in as little as 128 bytes of memory. 3 voice prints are required for an individual to "enroll" brining the total storage requirement between 512 Bytes and 1K per user with no additional reference database needed.

The original implementation of the keypad voice lock was based upon Motorola MC68HC11 processor. The circuit was designed to be mounted into a standard double gang box (4"x4"x3"). The input devices consisted of two keypads, one inside the secure space, and one outside. The user input also consisted of a small microphone mounted behind each keypad. There was only a red and green light provided for user feedback. The internal keypad was used for managing users voice prints, and Pins The external keypad could only send back key presses and audio information to the circuit board which was also mounted inside the secure area. The text display and audio used in this demonstration have been added for convenience.

Advances in PC technology have allowed our process to run without the aid of special dedicated hardware. Once the capture device send the verification sample to our smart card, for example, we can verify a human voice in only a few seconds, on a fully self contained 8K smart card running at 3.57 MHz

Furthermore, advances in micro chip technology allow this process to be integrated into many common consumer electronics devices for about 10 US$ cost of goods addition to an existing product (in production quantities). This cost of goods estimate includes our "VSS Certified" components and our proprietary firmware platform which drives the application.

Most "Internet" ($10-$20 stick type) desktop or monitor mounted microphones typically yield poor spectral information or low volume making them unsuitable for use with voice authentication applications. The algorithm has been tuned to work with a specific spectral response curve that many "Internet" microphones distort, losing valuable high frequency data. While the system may appear to be functioning properly, an impostor has a much greater chance of success at cracking voice prints created with inferior microphones..

We recommend using the Andrea NC-50 noise canceling headset for this PC based demonstration. Speak about 1/2 to 1" away from the microphone at a 90 degree angle to avoid breathing into the microphone.

The Andrea NC-50 Noise Cancelling Headset can be purchased from our eBay store, or directly from this page by clicking the button below.

For embedded hardware applications, we have selected a compact, VERY high-quality microphone which operates on 2-10 VDC current. These microphones are made in the USA and exhibit superior frequency characteristics. They are unique since they provide a full spectrum of speech even when the distance from the microphone is increased. They are only about the size of an LED and fit just about anywhere. We have yet to find a better microphone at any price.

Be sure to calibrate your microphone with a suitable audio application. Use a program such as Microsoft's Net Meeting™ , which contains an audio tuning wizard for setting up your microphone. You can obtain Internet Explorer and Net Meeting from the Microsoft web site.

For this demonstration to perform optimally, you'll need a microphone capable of capturing clear vowel sounds without clipping, distortion, or significant background noise. We also have hardware based demonstrations available on our custom hardware prototypes page.

Voice Security Systems Inc.
Biometric Technology Licensing
24591 Seth Circle, Dana Point, CA 92629
(949) 493-4030 tel (949) 388-7714 fax
E mail: info@voice-security.com
Privacy Notice | Conditions of Use
Power Voice®, FreeKey® and Voice Protect® are registered trademarks of Voice Security Systems Inc.
Copyright © 2004 Voice Security Systems Inc.
All Rights Reserved.