Biometric fusion-based security meets deep learning for smartphones & tablets
One of the hot areas in embedded space is that of biometric identification technologies, such as speech identification and facial recognition. Using just one of these techniques in isolation can result in an unacceptable level of "false negatives" (blocking access to the right person) and "false positives" (granting access to the wrong person). However, using both of these techniques together -- referred to as "biometric fusion" -- can yield extremely high levels of accuracy.
Sensory Inc. specializes in developing cutting-edge speech and vision recognition, identification, and verification technologies. Around the middle of 2014, Sensory introduced its first generation of TrulySecure, which combines speech identification and facial recognition to give a state-of-the-art security solution (see Voice & Face Unlock Smartphones & Tablets; also, this video offers a good overview).
Do you recall my recent column about neural networks and deep learning (see CEVA Accelerates Deep Neural Networks)? In the not-so-distant past, we used to be happy to be able to get an artificial neural network with just a few layers to work. More recently, it's become possible to train "deeper" networks containing 30 or more layers, and these have become known as "deep learning" neural networks.
As I said in that column: "Deep learning neural networks could prove beneficial for a wide variety of tasks, including object recognition, vision analytics, advanced driver assistance systems (ADAS), all the way up to artificial intelligence (AI)."
The reason I'm waffling on about this here is that the folks at Sensory have just announced the availability of TrulySecure 2.0 -- the next generation of FIDO-certified multimodal biometric authentication technology. In addition to leveraging Sensory's expertise in speech processing, computer vision to seamlessly fuse the two biometric modalities of face and voice, TrulySecure 2.0 incorporates next-generation neural networks and deep learning features to offer the most user-friendly, accurate, and secure multimodal authentication technology ideal for applications ranging from device unlock and application login to mobile payments.
Personally, I think this is a taste of a future that's racing towards us as we speak -- it won’t be long before things like text passwords and even fingerprint authentication are regarded as being "So early 21st Century, my dear!" I don’t know about you, but not being forced to create and remember yet another password would make me a very happy man indeed.