The development of Automatic Speech Recognition techniques continues to accelerate. Already an established technology, Automatic Speech Recognition is growing by leaps and bounds each year, especially as artificial intelligence contributes to evolution. A crucial building block of artificial intelligence is deep learning.
What is Deep Learning?
Deep learning refers to the process of a computer model learning how to do classification tasks by example, directly from audio, text, or images. These models are trained using very large sets of data and neural network topologies with many hidden layers, to which the word “deep” refers. Deep Neural Networks can achieve state-of-the-art performance in many different fields, even exceeding human-level performance on some of them.
What are Neural Networks?
More specifically, neural networks are a series of algorithms, whose job it is to identify relationships within a set of data, a process that simulates the way a human brain identifies underlying connections. When it comes to speech technology, neural networks enable us to push the limits of speech recognition.
Which Neural Network for Automatic Speech Recognition?
Deep Neural Networks are transforming the way humans interact, playing an important role in the technological revolution of artificial intelligence. At LumenVox, our Research and Development team is currently utilizing Time Depth Separable Convolutional Neural Networks (TDS CNN).
Convolutional Neural Networks are advantageous for a few reasons: They are computationally efficient, making them highly useful for mobile applications, and they have fewer knobs to toy with, fewer parameters to adjust. That means LumenVox customers get an ASR engine with greater speech recognition accuracy without requiring more compute performance, encouraging greater efficiency and performance.
LumenVox’ deep learning technology is applied to many of our technologies, including Automatic Speech Recognizer, Natural Language Processing, and Voice Biometrics. To learn more about our comprehensive stack, or to take an even deeper dive into deep learning, contact us today!
LumenVox Luminaries is a podcast that broadcasts thought leadership pieces on the subject of voice technology. This episode features Dr. Clive Summerfield, LumenVox Managing Director of EMEA-ANZ, discussing a case for active voice biometrics with his perspective on the specific benefits of Active Voice Biometrics.
I’m Clive Summerfield, Managing Director for LumenVox EMEA [Europe, Middle East, Africa] Australia and New Zealand, and I’m here to talk about Active Voice Biometrics. Many of you may have actually heard about Voice Biometrics, and I’m here to talk about the particular benefits of Active Voice Biometrics as opposed to Passive Voice Biometrics.
Q: What is Voice Biometrics?
Voice biometrics is a technology that allows a system to authenticate the identity of a person from an analysis of their voice. Like your fingerprint, your voice is unique. And so, like your fingerprint voice can be used as a very powerful technology for authenticating the identity of the speaker: you actually are who you say you are. And that’s a literal statement; the voice and the sound wave that emanates from your lips and your nostrils contains within it an acoustic signature that is unique to that individual speaker. And that in a nutshell is what Voice Biometrics is all about.
Q: What’s the difference between Active and Passive Voice Biometrics?
Voice biometrics comes in two flavors, essentially. There’s Active and Passive Voice Biometrics. Active is where a system asks you a specific question, and you have to answer that question with the correct voice. The obvious one here is a phrase like “My voice is my password,” but it can be anything—it can be your name; it can be your date of birth; it can be your address, your zip code or here in the UK your post code. It could be an account number or your telephone number. So this is where Active Voice Biometrics is Actively verifying your identity from a phrase that you’re actually saying. In Active Voice Biometrics you have to say the correct phrase with the correct voice. Now Passive on the other hand is a technology that sits in the background and listens to a conversation. Passive Voice Biometrics is, as the name implies, passively listening to a conversation and just understanding the voice characteristics of the speakers to recognize who the speakers are.
Q: Where is Active Voice Biometrics most appropriate?
The ideal application for Active Voice Biometrics is in telephone self-service and particularly for authenticating identity of speakers using IVR systems, particularly self-service applications—such as banking applications, government applications, applications for retail and telecommunications services. So in those applications the IVR prompts you for a piece of information. “Please say your telephone number.” And you have to say the correct telephone number with the correct voice in order to positively authenticate your identity, giving you a very strong level of surety that the speaker is the account holder and not an imposter trying to break into your accounts.
I feel like the big application for Active Voice Biometrics has been in telephone-based password reset. And there are numerous examples of Active Voice Biometrics being used for telephone password reset applications, and those are principally in help desks and internal-facing employee helpdesk applications and services. But Active Voice Biometrics is far more than just a password reset application; it is also a password replacement technology where voice can be used instead of passwords for many applications. These include IVR telephone self- service chat bots and increasingly online services and IOT devices.
Q: What about the future of Active?
Well, there is good news in that voice interactions are growing in the world, I mean, despite the demise of telephone call centers, the growth in voice communications is growing exponentially at the moment. According to Gartner, 30% of all searches are now voice-driven which tells you where voice communications is going in the future. So the digital channel is where all the action is actually going to be in the near term.
And whilst Active still has a very important role to play in telephone services, the future of Active is actually in the digital channel. Now digital channels are almost by definition driven by phrases and single phrases, which is where Active has very strong application over and above Passive. So Active is the only technology in the digital channel that is applicable for things like second-factor authentication. I like to think that Voice Biometrics and Active Voice Biometrics is actually the world’s best second-factor authentication which can be used to augment things like PINs and passwords that are traditionally stored up now in browsers. Active Voice Biometrics can be very effectively applied in digital channels and particularly in second factor authentication for browsers channels.
One of the big applications for Active Voice Biometrics is capturing voice through the browser as somebody is accessing a secure website such as Internet banking and other services of that nature, to positively authenticate that the person who is actually the account holder. This provides a much stronger security credential than PINs or passwords on their own. And Active is unique in so far that it’s the only technology that can harmonize the authentication between the emerging digital channels and the legacy telephone channels. You can use the same phrase in a digital channel as you can in a telephone channel. There’s no other biometric technology that provides this type of flexibility which means that organizations can harmonize their authentication strategies across all their customer service channels thereby providing a much more effective customer service experience.
There are three key issues on credit unions’ minds right now: Fraud, operations, and member experience. Fraud is a rising issue due to uncertain economic times. Operational efficiencies are a high priority because the contact center is now being inundated with calls—time, more than ever, is money. Member experience is always at the top of the priority list because it inspires every roadmap.
Is there a solution that can address all three?
There is. Passive Voice Biometric Authentication effortlessly authenticates members over the phone during the natural course of conversation with an agent. Members register their voiceprint, and on subsequent calls, their voice is matched to this unique identifier, giving the agent a green light when the identity claim is genuine and a red light when a fraudster is attempting an account takeover. This solution addresses these areas with these three key strategies for credit unions to tackle the following:
Passive voice biometrics, including fraudster detection, is a layer of security that seamlessly works with humans to prevent mistakes that often go unnoticed today. When a call comes in, the caller’s voice is compared to the voiceprints of known fraudsters. The agent is alerted and coached from there to take evasive action.
Reduce Agent Handle Times. Agents dread the first few minutes of every call. With LumenVox Passive Voice Biometric Authentication they can start solving the caller’s problem with confidence, verifying identity within seconds, not minutes. LumenVox Passive Voice Biometric Authentication results in reducing AHT by 30 – 60 seconds per call. This drives cost out of the contact center and improves your bottom line.
According to a study by Salesforce, 84% of customers say the experience a company provides is as important as its products or services. Voice biometrics enable your credit union to give your members a positive authentication experience, from the very beginning of the call. With LumenVox Passive Voice Biometric Authentication the member can be authenticated while speaking comfortably and naturally to an agent, reducing the need for a long line of interrogative questioning. Contact centers with passive voice biometrics start by asking their customers “How are you today?” instead of “Who are you, and are you who you say you are?”
Passive Voice Biometric Authentication is available today. Interested in learning more? Watch the short video on this page and contact us today!
It was recently announced that a biometric data privacy suit implicated the voice biometrics company, Pindrop. According to the plaintiffs, informed consent requirements were not followed when collecting, storing, and utilizing voiceprints to authenticate customers for Amazon Web Services, which the plaintiffs claim violates the Illinois Biometric Information Privacy Act (BIPA).
This scenario is exactly what LumenVox works hard to avoid. Since LumenVox is headquartered in California (where the biggest giants of customer data work, such as Facebook and Google) and operates internationally, we understand that customers demand privacy. There are now stringent laws in place, such as the California Consumer Privacy Act and GDPR, to make sure that just as a company’s technology evolves, so must the strategy to correctly collect, handle and store sensitive consumer data.
When it comes to any biometrics modality, informed consent is key. LumenVox advises our customers and partners to include a very clear information flow to gain informed consent for both active and passive voice biometrics. As a voice biometrics provider, education, training, and guidance are three things we pride ourselves on as we work closely with our clients on voice biometric deployments.
Biometric solutions have been around for decades now, but understandably consumers and businesses still have their reservations. What’s particularly comforting about voice biometrics is the way voiceprints are stored – with Personal Identifiable Information (PII) decoupled from the voiceprint. This means a consumer’s voiceprint is matched to a string of 0-s and 1-s, numbers that would mean nothing to a malicious actor.
This news inevitably raises questions: Is voice biometrics safe? Yes. Can it be used to protect customer data? Yes. Is informed consent key when utilizing it? Absolutely. At LumenVox, we strive to put customers first, every step of the way.
Voice biometrics provides a robust layer of security for contact centers, using the physical and behavioral characteristics of a voice to verify claimed identities. But just as businesses are adopting new voice biometric technology, fraudsters continue to advance and adapt techniques to take over customer accounts.
There has been concern expressed that voice biometrics is vulnerable to a certain level of sophisticated fraud. This concern rises with the evolving technology surrounding synthetic voices. A synthetic voice is a collaboration between a human being and a machine to produce an artificial version of human speech.
The truth remains: Fraudsters can indeed try to elude biometrics authentication using synthetic voices. And yet this truth doesn’t scare us, and here’s why:
Synthetic speech leaves evidence, a trail. It generates specific artifacts which makes this type of speech detectable.
Even in the best of cases, the synthetic voice doesn’t sound entirely human. With passive authentication, there is always an agent who acts as a spoofing detector. It’s very easy for humans to correctly distinguish synthetic speech from real human speech. As humans, we know when something sounds off.
The reality is fraudsters follow the path of least resistance; they want to work the fewest number of hours for the biggest amount of payoff. The creation of a synthetic voice requires a very high level of effort. The best deep fake or synthetic voice technology currently needs a substantial amount of audio data to train the voice. Getting this in normal circumstances is difficult, and yet not impossible. While there are tools that exist which generate synthetic speech, they require a certain level of skill/ expertise to utilize. These combined factors act as a natural deterrent.
Having made those points there is another truth which remains: All the detection, human intervention, and natural deterrents in the world do not necessarily equate to peace of mind. At LumenVox, we encourage the use of voice biometrics alongside traditional authentication methods, including PINs, passwords, security questions, so that customers are protected with multiple layers of security.
To learn more about the LumenVox Security Suite, visit our products page here.
Did you know that during normal conditions, approximately 30% of Help Desk calls are for routine password resets? With the majority of employees currently working from home, that percentage is set to increase exponentially.
Businesses often find themselves stuck with a password reset process that is costly, time-intensive, insecure, and anything but user–friendly. It’s a pain point that is often overlooked. LumenVox Password Reset provides a fully self-service automated password reset function which requires little effort from the user with only seconds to complete: To reset a password, your employee/user is prompted to speak a short passphrase, and through voice biometrics the user’s claimed identity is validated and the password is reset. The solution is highly secure, utilizing multifactor authentication modalities, including:
Biometrics: This uses something you are – for our solution, it’syour voice.
Knowledge-Based Authentication: This is something you know, such as the name of your first pet, etc.
Push Notification: This is something you have (mobile device, email…).
Computer Certificate: This is something you have and know.
LumenVox Password Reset enables a seamless authentication process that radically improves security and convenience for your remote workforce by reducing labor costs through the automation of Password Resets; providing access to additional custom identity validation methods - works with directory systems and applications such as Active Directory, LDAP, SAP, Oracle and more; and providing your users with multiple options: Phone, Mobile App, Windows Pre-Login Web Portal.
The solution providing employees 24/7/365 accessibility to unlock their accounts, so they can self-heal, even when the Help Desk is closed. It also simplifies integration with directory platforms, IVRs and Help Desk applications and safeguards user information.