Herself’s Artificial Intelligence

Face recognition technology is now far enough along and mainstream enough that is it available to the masses. At least if you buy your notebook from NEC.

NEC has launched two new series of laptops with a unique security feature called “face pass” — or, in Japanese, “kao pass.”
The LaVie C and LaVie L series will both include the new facial recognition software, which enables only a programmed user to log on to the computer.NEC’s software, called “NeoFace,” is a biometric system that uses a combination of eye zone extraction and facial recognition to identify the computer’s user. To program the system, a user sets up a profile with three photographs of their face. Then when a user tries to log on, an integrated 2.0 megapixel camera scans their facial characteristics.

The NeoFace system then uses a matching procedure to determine the identity of the user. NEC says that the system performs accurate matching even when people wear glasses and hats, have different haircuts or facial hair, and show different facial expressions. The ability to distinguish between identical twins is still speculative.

“NeoFace uses a technology called ‘adaptive region mixed matching,’ which focuses on ‘segment regions’ with a high degree of similarity for matching,” explained Atsushi Sato, a head researcher at NEC. “Other makers’ products make judgments based on a number of combined characteristics, such as the distance between the eyes and the nose, or the nose and the mouth. But this creates a problem, because if even one of these segments is missing, the accuracy drops dramatically.

“In contrast, NeoFace divides the input image and the registered image into small segments, and focused only on the segments that are highly similar,” he continued. “This enables the system to achieve higher authentication accuracy than out competitors’ products, even if a part of the subject’s face is hidden, for example by a mask or sunglasses.”
‘Face pass’ is latest security system for NEC Laptops

See also:
NEC NeoFace
Advances in Face Detection and Recognition Technologies ( pdf)

Millions of inventions pass quietly through the U.S. patent office each year. Patent No. 7,033,406 did, too, until energy insiders spotted six words in the filing that sounded like a death knell for the internal combustion engine.

An Austin-based startup called EEStor promised “technologies for replacement of electrochemical batteries,” meaning a motorist could plug in a car for five minutes and drive 500 miles roundtrip between Dallas and Houston without gasoline.By contrast, some plug-in hybrids on the horizon would require motorists to charge their cars in a wall outlet overnight and promise only 50 miles of gasoline-free commute. And the popular hybrids on the road today still depend heavily on fossil fuels.

“It’s a paradigm shift,” said Ian Clifford, chief executive of Toronto-based ZENN Motor Co., which has licensed EEStor’s invention. “The Achilles’ heel to the electric car industry has been energy storage. By all rights, this would make internal combustion engines unnecessary.” . . . Texas startup says it has batteries beat

Ah, cool but what does this have to do with artificial intelligence you ask? The biggest holdup in robotics right now is just that — battery life. Smaller, longer lasting batteries would allow for huge advances in the field of robotics. Let’s hope this technology proves itself.

Stanford School of Engineering, AI/Machine Learning

Video Lectures.net has flash formatted video lectures. Right now there are many on computer science including Data Mining, Data Visualization, Fuzzy Logic, Image Analysis, Information Extraction, Information Retrieval, Machine Learning ( 326 lectures! ), Robotics, The Semantic Web and Text Mining.

Or you can try one of the many MIT Computer Courses These usually list the text book, have a syllabus, lecture notes, sometimes videos and homework and test problems and solutions.

Also there is MIT’s Introduction to Neural Networks

And Introduction to Machine Learning a PDF textbook draft you can download.

There are also many documents on Machine Learning at Scribd.

UCI Machine Learning Repository has over 160 data sets for you to use to test and develop your AI.