recognition
Latest
Automated English visa test struggles to understand English
An Irish veterinarian's application for an Australian visa has been rejected after she failed to pass an automated English proficiency test, despite completing it in her native language of... English. Louise Kennedy, who has two degrees (both obtained in English), wanted to apply for permanent residency in the country on the grounds of her job, which is classed as a shortage profession. Despite acing the reading and writing parts of the test she didn't score highly enough on oral fluency, as it seems the machines couldn't understand her accent. The Pearson Test of English (PTE) Academic is an automated system that asks applicants a number of questions and records their vocal responses which are analyzed and scored. The Australian government demands a score of at least 79 points. Kennedy scored 74.
Shazam finally syncs matched tracks across devices
Shazam has a lot more competition from Siri, Cortana and other services, but the music recognition service has tried to stay on top with Google Now integration, faster recognition and other new tricks. The company has just added a feature that seems indispensable, but somehow wasn't available before: syncing across devices. As long as you're logged in to your account, it remembers any songs you identify and lets you see them from the desktop, iOS or other devices.
Computers can categorize buildings into architectural styles
Even if you've never heard of "Byzantine," you can probably tell a Byzantine church from a Gothic one. Judging style differences is nearly impossible for a computer, however, and researchers from the University of Massachusetts want to fill in that gap. They used geometric matching, crowdsourcing and machine learning to teach an algorithm how to spot similar styles in buildings, furniture and other objects. That's something that could be incredibly useful for historians with mountains of photo archives, or game designers who need to auto-fill a level with historically accurate furniture.
Apple buys tech that could take Siri offline
Apple has sort-of-confirmed that it recently snapped up another small company, called Novauris. The firm specializes in speech recognition and has historical ties to the core technology and patents underpinning Siri. TechCrunch reports that Novauris' experts are already working inside Apple to improve its voice assistant, but no one really knows exactly what they're up to. One of Novauris's big strengths has been locally processed recognition, which doesn't rely on distant servers, so it's possible that Apple wants Siri to accomplish more without a data connection. (Apple's current Siri partner, Nuance, can also do offline processing, but Apple hasn't been able to bring that technology in-house.) We're just speculating, of course, but this is a function that no voice assistant has really mastered so far (although others are definitely working on it), and it's even more important now that iOS is getting into the car.
BYU image algorithm can recognize objects without any human help
Even the smartest object recognition systems tend to require at least some human input to be effective, even if it's just to get the ball rolling. Not a new system from Brigham Young University, however. A team led by Dah-Jye Lee has built a genetic algorithm that decides which features are important all on its own. The code doesn't need to reset whenever it looks for a new object, and it's accurate to the point where it can reliably pick out subtle differences -- different varieties of fish, for instance. There's no word on just when we might see this algorithm reach the real world, but Lee believes that it could spot invasive species and manufacturing defects without requiring constant human oversight. Let's just hope it doesn't decide that we're the invasive species.
NEC outs cloud computing facial recognition service for merchants
NEC has launched a $880 per month service in Japan that lets merchants profile customers using just a PC and video camera. The system uses facial recognition powered by the company's cloud computing service to estimate the gender and age of clients, along with the frequency of their shopping expeditions across multiple locations. The firm developed the "NeoFace" tracking software in-house, claiming it was the highest ranked facial recognition system in NIST and that it plans to use it for other services like "intruder surveillance" in the future. NEC added that face data is encrypted so it can't be "inadvertently disclosed," and is strictly to help retailers fine-tune their marketing strategies. After watching the system pick off face after face in the video after the break, we just hope it doesn't go rogue.
FBI to roll out $1 billion public facial recognition system in 2014, will be on to your evildoing everywhere
They're watching you -- or at least will be in a couple of years. That's when the FBI is gearing up for a nationwide launch of a $1 billion project designed to identify people of interest, according to the New Scientist. Dubbed the Next Generation Identification (NGI) program, the high-tech endeavor uses biometric data such as DNA analysis, iris scans and voice identification to track down folks with a criminal history. The FBI also plans to take NGI on the road literally by using public cameras to pick faces from the crowd and cross check them with its national repository of images. Let's just say this facial technology isn't going to be used for lighthearted Japanese vocaloid hijinks or unlocking your electronic device. The use and scope of NGI, which kicked off a pilot program in February, will likely be questioned not just by black helicopter watchers but privacy advocates as well. Facial recognition has certainly been a touchy issue in privacy circles -- something Facebook learned firsthand in Germany. Meanwhile, the Electronic Frontier Foundation is already raising concerns about innocent civilians being mixed up or included in the database. Naturally, the FBI claims that the NGI program is in compliance with the U.S. Privacy Act. On the positive side, at least they didn't name it the Genetic Lifeform and Disk Operating System.
Google lands patent for automatic object recognition in videos, leaves no stone untagged
Google has already been working on patents that could pick out faces and song melodies in our YouTube clips. Now, it might just have the ultimate tool: the technique in a just-granted patent could pick out objects in a video, whether they're living or not. Instead of asking the creator to label objects every time, Google proposes using a database of "feature vectors" such as color, movement, shape and texture to automatically identify subjects in the frame through their common traits -- a cat's ears and fast movement would separate it from the ball of yarn it's attacking, for example. Movie makers themselves could provide a lot of the underlying material just by naming and tagging enough of their clips, with the more accurate labels helping to separate the wheat from the chaff if an automated visual ranking system falls short. The one mystery is what Google plans to do with its newfound observational skills, if anything, although the most logical step would be to fill in YouTube keywords without any user intervention -- a potential time-saver when we're uploading that twelfth consecutive pet video.
EyeRing finger-mounted connected cam captures signs and dollar bills, identifies them with OCR (hands-on)
Ready to swap that diamond for a finger-mounted camera with a built-in trigger and Bluetooth connectivity? If it could help identify otherwise indistinguishable objects, you might just consider it. The MIT Media Lab's EyeRing project was designed with an assistive focus in mind, helping visually disabled persons read signs or identify currency, for example, while also serving to assist children during the tedious process of learning to read. Instead of hunting for a grownup to translate text into speech, a young student could direct EyeRing at words on a page, hit the shutter release, and receive a verbal response from a Bluetooth-connected device, such as a smartphone or tablet. EyeRing could be useful for other individuals as well, serving as an ever-ready imaging device that enables you to capture pictures or documents with ease, transmitting them automatically to a smartphone, then on to a media sharing site or a server. We peeked at EyeRing during our visit to the MIT Media Lab this week, and while the device is buggy at best in its current state, we can definitely see how it could fit into the lives of people unable to read posted signs, text on a page or the monetary value of a currency note. We had an opportunity to see several iterations of the device, which has come quite a long way in recent months, as you'll notice in the gallery below. The demo, which like many at the Lab includes a Samsung Epic 4G, transmits images from the ring to the smartphone, where text is highlighted and read aloud using a custom app. Snapping the text "ring," it took a dozen or so attempts before the rig correctly read the word aloud, but considering that we've seen much more accurate OCR implementations, it's reasonable to expect a more advanced version of the software to make its way out once the hardware is a bit more polished -- at this stage, EyeRing is more about the device itself, which had some issues of its own maintaining a link to the phone. You can get a feel for how the whole package works in the video after the break, which required quite a few takes before we were able to capture an accurate reading.
LookTel releases recognizer app for the visually disabled
LookTel, which came out with the Money Reader app last year that helps the visually impaired accurately count money, has come out with a new app using the same technology to help people with vision problems to identify common objects. LookTel Recognizer allows users to magnify items and store them to a database for later reference. You can use this for groceries, credit cards, ID, medicine or anything you can think of. Snap a picture with an iPhone, add an audio message, then save it to the database. When someone who is visually impaired uses the app, they can point the iPhone at an object. If it's in the database, it'll be recognized, and the user is informed what the item is using the prerecorded audio message. LookTel Recognizer is out now for $9.99, and you can see it in action in the video below.
Mountain Lion may be able to scan a barcode, install an app
It's early days with the developer preview of OS X 10.8 Mountain Lion, but there are already a few interesting tidbits peeking out around the corners of the developer NDA. We've heard that the system includes a new 'CoreRecognition' framework and a 'CRCodeRedeemer' class, both of which appear to be built for turning barcodes into app licenses. While marketers can use a 2D barcode now to lead customers to their website or app, that doesn't necessarily translate to an app download or purchase. If this framework operates as it appears, that won't be a problem in Mountain Lion. Specially-crafted barcodes would let you simply hold up a card to your Mac's camera to download a prepaid app from the Mac App Store (or from iTunes, possibly). That's a new level of convenience for developers who want to give away copies of their apps at face-to-face events. This technology would be even more interesting if Apple's 2011 patent application for 'digital handshakes' ever sees daylight. With barcodes or data embeds printed on your devices with invisible ink, you could log into your Mac just by waving your iPhone at it -- which would require that your Mac have some barcode-reading savvy.
Apple applies for facial recognition patent, wants to let iDevices get to know you better
Your momma always said your handsome mug would take you places. Now it might allow you to access your iPad. An Apple patent application released today describes a facial recognition system that requires minimal computing power, and works whether you're indoors or out -- we don't use our tablets and phones in a photo booth, after all. The technology works by comparing a current image of your mug to a reference model user profile made using "high information" portions of the human face, like eyes and mouths. Translation: it'll take a picture, compare it against the pictures associated with various user accounts on the device and decide if the two images are similar enough to grant you access. Because this is just an application, it's safe to say we won't be seeing this kind of facial recognition in iOS anytime soon, but let's hope it works better than the ICS version if it does. Update: An important thing to note is that Apple applied for this patent long before Android's Face Unlock debuted a few months back. The paperwork was first submitted on June 29th, 2010 -- it's just now being disclosed to the public.
Nuance Dragon Dictate 2.5 for Mac review
Voice recognition. Or, more specifically, speech recognition. It's one of those technological wonders that we all seem to take for granted, while simultaneously throwing laughter its way for not being nearly sophisticated enough. Anyone that's used an early generation Ford SYNC system -- or pretty much any vehicular voice command system -- knows exactly what we're getting at. While processing speeds and user interfaces have made great strides in the past handful of years, voice recognition has managed to continually disappoint. It's not that things aren't improving, it's just that they aren't improving at the same rate as the hardware and software surrounding them. Even today, most new automobiles have to be spoken to loudly, pointedly and directly, and even then it's a crapshoot as to whether or not your command will be recognized and acted upon. For as much as we complain, we totally get it. Teaching a computer program how to recognize, understand and act upon the movement of human vocal chords is a Herculean task. Throw in nearly unlimited amounts of dialect and regional variation with even a single language, and it's a wonder that programs such as Nuance's Dragon Dictate even exist. Teaching a vehicle how to route calls, adjust volume and tweak a radio station is one thing, but having a program that turns actual speech into presentable documents requires a heightened level of accuracy. The newest build of Dragon Dictate for Mac (v2.5) allows users to seamlessly combine dictation with mouse and keyboard input in Microsoft Word 2011; it also gives yappers the ability to more finely control how Dragon formats text such as dates, times, numbers and addresses, while a free iOS app turns your iPhone, iPad or iPod touch into a wireless microphone. We recently pushed our preconceived notions about this stuff aside in order to spend a solid week relying on our voice instead of our fingertips -- read on to see how it turned on. %Gallery-132799%
Four Grammy Awards renamed to include 'video games music,' underline its growing importance
Video games have never had a problem sitting alongside movie DVDs and music CDs (back when such things were distributed physically) in stores, so it's frankly overdue to see them pop up in a mainstream awards show like the Grammys. The US Recording Academy has finally deigned to address video games and their aural landscapes as a separate class of entertainment, and has now amended four of its awards to spell that out. What was formerly known as "Motion, Television, or Other Visual Media" is now reclassified as "Motion, Television, Video Games Music, or Other Visual Media," leading to there now being four awards that explicitly recognize excellence in video game music scores. Guess that was inevitable after Christopher Tin's Baba Yetu won a Grammy this February, ostensibly because of its inclusion as one of the marquee songs on Civilization 4, but in a category entirely unrelated to gaming. Skip past the break to hear it for yourself, along with a couple of other favorites of ours.
Microsoft's OneVision Video Recognizer can detect, identify, and track your face on video... so smile!
Here's your classic case of "just because you can, doesn't mean you should." Microsoft's Innovation Labs have just demonstrated a OneVision Video Recognizer algorithm that's powerful enough to perform face detection duties on a running video feed. It can recognize and track humanoid visages even while they're moving, accept tags that allow auto-identification of people as they enter the frame, and can ultimately lead to some highly sophisticated video editing and indexing via its automated information gathering. Of course, it's that very ease with which it can keep a watchful eye on everyone that has us feeling uneasy right now, but what are you gonna do? Watch the video after the break, that's what.
iPhone app helps the blind ID currency
Identifying different denominations of dollar bills has long been a problem for the sight-impaired in the US. A US$1 bill is the same size and shape as a $100 bill, making them hard to differentiate. Many use techniques like folding different bills into different shapes, but this doesn't help when receiving change in a store. LookTel now has an app called Money Reader for the iPhone 4/3GS and fourth-generation iPod touch which will check bills and speak their values out loud. LookTel says its Money Reader can use the iPhone's camera to "read" currency and speak its value aloud in real time, and the app doesn't need an internet connection. It currently recognises $1, $2, $5, $10, $20, $50 and $100 bills, and it costs $2 from the App Store. The Treasury Department was ordered last year to change US currency to make it easier for the blind and partially-sighted to identify it -- other countries make their notes different sizes and shapes to help with this -- but until then, this could be a big help.
Papershow for Teachers introduced at Macworld 2011
Papershow made a big splash last year at Macworld during its introduction, and since then has apparently sold fairly well in business circles (the product is now being sold in Staples as well). This year, Papershow is introducing a new edition styled for education applications, called Papershow for Teachers, that features not only a few upgrades to the functionality, but a few targeted extras meant specifically for the classroom. Just like the previous edition, Papershow for Teachers has a special pen (with a camera built into it) and a Bluetooth-enabled USB drive that hooks right into your Mac with all the necessary software pre-installed on it. The pen's camera can only be used with special paper (a pad is included with the package, and you can, of course, buy more) that has a series of dots on it to tell the app where and when you're drawing on the screen. So very easily, anything written on the special paper gets translated up to the screen. That's the basic functionality, but a printed toolbar on the paper adds a lot more. By "pressing buttons" on the printed toolbar, the pen can change colors, change shapes, or even do things like pull up graphics or draw shapes on screen. The new teacher edition includes an extended toolbar with even more functions -- teachers can now pull up a virtual calculator on screen, and even do things like draw an angle and then ask the software to calculate exactly what it is. Any images created with the app can be exported out to email or PDF, so teachers can share notes with their students right away.
Apple granted patent for handheld that recognizes your hands
You could probably fill a book with Apple patent applications that never amounted to anything, but here's one that's at least been granted -- a "handheld device" that uses capacitive sensors to recognize your identity just by the way you hold it, and subsequently personalize the device's buttons and settings to your hand based on your user profile. That's all that's actually been patented here, but the general idea is a little more grand -- you could theoretically grip a handheld with either hand, and it would automatically generate "button zones" under each finger using sensors (or disappearing buttons, perhaps) baked right into the chassis, making cries of "you're holding it wrong" hopefully fade into obscurity. Still, it's not the first time we've seen Cupertino reaching for digit recognition and we're sure it won't be the last, so we'll just file away our enthusiasm until or unless Jobs announces it on stage.
Shazam updates for iOS 4.0, adds subscription model for Encore features
The great Shazam app, which I'm still using after downloading it to my iPhone on day one, has released another update, bringing the whole suite of apps (including the paid Shazam Encore and the Shazam (RED) version) to version 3.3. This one updates the app for iOS 4.0 and includes Retina Display graphics and support for fast app switching. Version 3.3 also adds some setting customization options like "tag on startup," improved artist info, a new Discover tab with tag charts and preview clips for the top songs tagged in 20 different countries, and quick shortcut icons for sending the music off to iTunes or to Facebook and Twitter. Additionally, the free Shazam app now offers a subscription model upgrade to the paid version, which allows for as many tags as you want, extra features like "Drive and Tag," and more information about the songs you hear with the app. For a limited price of $2.99, you can get access to Shazam Encore for an entire year, or for $4.99 (the price of the paid app on the App Store), you can get a "lifetime subscription." Clearly Shazam wants to offer something a little cheaper for customers who aren't sure if they'll use the app that much, but honestly, if you want the pro options, the Encore version is worth the $5. On the other hand, all I've ever had is the free version, and for the occasional "hear a song on the radio or in a store and wonder what the heck it is," Shazam has served me quite well. Good to know it'll now look and run better on the iPhone 4.
EA Sports honors 'Game Changers' in its online communities
Over the weekend, EA Sports announced a new community influencers program that it's calling "Game Changers" -- the idea is that the publisher will choose big EA sports fans who use their online community know-how "to improve the experience and overall growth of the EA SPORTS Community." Presumably, you have to yell "EA Sports" whenever you say it, which would be why the name is in all caps. The chosen few will get early access to EA games, as well as represent the community while giving feedback on future releases. They've picked nine guys (no girls) so far, all of whom are active in the various EA communities and forums. They've also posted a video of the program's kickoff, and while we were totally going to make fun of it for sounding like a cliched sports team ("It's the dream of a lifetime, we just go out there to do our best" and so on), that Anton B dude actually gets a little choked up at the end after being honored with Game Changer of the Month. So all right, Anton, we get it. This Game Changers thing may not mean too much to the rest of us (just like the Xbox Ambassador program, it's just a way for the company to connect directly with their community), but everybody's got to have their thing, and this is yours. Rock on for that.
