|
 |
artiNNet Technology artiNNet's patent-pending technology uses a proprietary family of optimum-entropy neural net algorithms to process biometric data at more than 1000 times the efficiency of today's prevailing algorithms. This quantum leap in efficiency enables, for the first time, robust biometric authentication on inexpensive, low-power integrated circuits. Our algorithms require just 8K of memory (for both logic and data), require only negligible power (2-3 milliamps per authentication), and process at speeds under 1 millisecond. For example, our fingerprint authentication algorithm need only perform 8000 multiplication and addition calculations per authentication, as compared to current industry-standard algorithms that must perform 250,000-500,000 calculations. These massive algorithms must be run on fast, expensive processors, which in turn require external control devices and rechargeable batteries. In contrast, our algorithms operate on any logic calculation chip with 8K of available memory. It is therefore easy to hardcode our algorithms into 16K smart chips, for instance, or port them in software to existing devices' standard microprocessors, such as the ARM7, ARM9, TI DSP, and so on. With such minimal memory, processor, and power requirements — and the resulting dramatically lower costs — artiNNet has not only made biometric authentication available to entirely new markets for just a few dollars per device, but we have also solved the privacy concerns that have hindered the industry until now. One Algorithm, Infinite Devices A unique feature of our core algorithm is that it is completely sensor-agnostic, platform neutral, and datatype-neutral, which means that it can be quickly and inexpensively adapted to virtually any device, whether the device uses fingerprint, voice, iris, retina, hand geometry, face recognition, or any other anatomical, physiological, or behavioral biometric data type. Combined with recent, dramatic industry innovations in thin-sensor technologies, artiNNet's algorithm makes possible full-fledged, 100% stand-alone fingerprint scanning and verification on devices as thin as 0.84 millimeters, which includes ISO-compliant smart card and ordinary magstripe credit cards. Technology features include:
One Algorithm, Infinite Applications Owing to the breadth of neural network applications, our algorithms have also been applied very successfully to many non-biometric problems, such as optimizing chemical manufacturing processes and digital signal processing of SONAR data in naval submarines. Potential applications of such predictive behavior models are limited only by one's imagination and range from sophisticated real-time financial analysis to artificially intelligent military devices. We encourage inquiries for any biometric application or non-biometric application that would benefit from neural network algorithms applied to low cost, low power, portable, or ultra-miniaturized devices. Click here to learn more about licensing our algorithms. Contact Us |
