Voice biometrics hit the headlines last year to the surprise of many and the fascination of others. While the terms voice and speech recognition was already widely known thanks to virtual assistants and other applications derived from the use of smartphones, biometrics presented a new way of using voice for different purposes.
Identity fraud and impersonation have shared space in the same conversations and forums as voice biometrics, as well as in those dedicated to customer experience and sales team productivity. Companies in sectors such as banking, financial services, telecommunications and utilities are already making use of this technology, which has proven to be reliable, secure and opportunity-generating.
Voice biometrics is a technology capable of recognizing and identifying users through the collection of sound data obtained from a person's voice. As a biometric identification factor, voice represents a physical trait that is just as differentiating as fingerprints, iris or face.
This technology is capable of creating a biometric sound pattern derived from a user's speech. The comparison of two sound samples and their differentiation is the basis of identification through this method. The voice is an irreplaceable, personal and non-transferable authentication factor. Also known as vocal biometry, voice biometry works on the vibrations and waves generated by a person when speaking and which are the product of a set of organs (mouth, teeth, tongue, throat, vocal cords, pharynx, oral-nasal cavity...) unique and specific to the individual. The result of the joint work of each gear of the organism is the unique vocal tone of each person.
It is common to wonder about the accuracy and reliability of this technique since we have all gone through stages where our voice has been modified either by a simple illness such as a cold or simply by the fact that we woke up a few minutes ago. Similarly, the passage of time and ageing modify our vocal tone.
The know-how of the most expert engineers and the use of the most avant-garde and powerful technology in the areas of artificial intelligence and machine learning provide a clear and reliable response to the challenges posed by identifying the voice of a user at very different times and under very different circumstances.
Proper training of the platform with sufficient voice samples in the recording phase is crucial to obtaining a quality voiceprint. Thus, the best voice biometrics technologies analyze at least 100 different parameters associated with our voice resulting in an absolute success rate.
The methods and technology are composed in a very similar way to the standardized and widespread facial recognition, which does the same by generating a mathematical pattern of the face.
For the training of voice biometrics systems, there are usually three voice samples requested from the user for the first time. In parallel, they are analyzed during the test phase and compared for verification with each other as well as with other samples to generate the differentiated identification pattern.
There are two types of voice biometrics for authentication. Closely related to the time at which the system is run, the type chosen will depend on the use case and the application in which the authentication is to take place:
This type of voice authentication prompts users to say a phrase or words aloud a certain number of times so that the system can create the voiceprint and associate it with their identity. Then, when the user wants to authenticate, they must recite a phone number, their account number or another specific phrase to match the voiceprint.
That is, the user must perform an action by actively participating by specifying a phrase that serves as a password. This type of biometric identification by voice has double security: recognising the vocal pattern and knowing a password or data.
The difference between active voice biometrics and passive voice biometrics lies in the user's participation as his or her voice is recognized. Here, it is not necessary for the user to say specific words or phrases, but just to speak and say anything for a few seconds.
The content of what is said is not relevant to consider authenticating you. In this method, the vectors and factors that interpret the sound information are maximized, making it a secure and infallible method of identification, as well as simple for the user. It does not matter the text or even the language in which you speak, with security controls that detect prerecorded voices.
Beyond unlocking a device, voice biometrics has found its place among the biometric identification techniques used to enroll, authenticate and verify customers and users in multiple use cases. Companies and businesses across all sectors and industries are integrating this technology, although the following areas of activity stand out:
Voice biometrics systems are integrated into the processes and operations of these areas through all kinds of channels. Like facial recognition, this technology is taking place in customer onboarding processes either through an app or a website. On the other hand, newer applications of it are integrating it into chats and virtual assistants available to customers on different platforms.
However, although they take place in all these channels, the use in videoconferences and call centers (VRI and sales agents or telemarketers) stands out. In this implementation, voice collection can take place before or during the contact with the commercial agent, depending on the use case.
As we have been discussing, identity verification is the backbone of the KYC (Know Your Customer) processes that occur during the digital onboarding of users. While the already standardized facial recognition is the norm and has an absolute success rate, voice biometrics is emerging to be part of KYC processes, either as a complement or as a unique identification factor.
After registration during onboarding, the voice pattern storage is used to authenticate the user when they log-in to customer platforms to access manage their products and services or manage their preferences. This is where voice biometrics is strong, as many users still make their queries over the phone.
Commercial agents can have prior information before attending to the user because the user interacted with his voice in the assistant prior to the referral to personalized attention. On the other hand, another form of implementation consists of recognizing the user in the first few seconds when he is asked what he wants to consult about, which is enough time to identify him by voice and display his file.
The creation of the biometric voice pattern can be created not only in the first registration and onboarding but can also be proposed to the user in any contact with him and inform him that it will be used to authenticate him in future contacts with more agility and ease for him and for the agents that will attend him.
On the other hand, like face, voice can act as a second authentication factor in Strong Customer Authentication (SCA) MFA strategies on any platform. The combination of face and voice would prevent SMSs that may incur SIM Swapping or forgotten credentials.
Both telephone and online transactions and payments are already implementing voice biometrics, for example, to authorize bank transfers. More disruptive applications are also proposed in the tourism and hotel area, such as online check-in with voice and even opening doors in hotel rooms, similar to what is already done with the face.
The advanced electronic signature is the most agile and versatile method of contracting and acceptance of conditions. The implementation of this tool in businesses in all industries and areas has brought huge benefits and triggered both growth and resource optimization.
Voice biometrics can be used in digital signatures to identify your users by voiceprint. Supported by the eIDAS (electronic IDentification, Authentication and trust Services) common framework, and endorsed by the most demanding auditors, it has full legal validity and is perfect for telco or energy sector contracting through call centers.
Voice biometrics is a new approach to improving the user experience when interacting with brands, platforms and businesses. The obligation to identify users in order to comply with KYC, AML or other regulatory requirements associated with industry standards has prompted technology companies to develop different authentication methods that have the least possible impact and avoid any type of friction.
It has proven to be as accurate as other biometric identification techniques and equally scalable and affordable. Its ease of integration and the availability of microphones in most devices have made it a very viable option today.
Voice biometrics has been the answer to agility in telephone contact. The TMO (Mean Time To Operate) has improved considerably thanks to the implementation of voice biometrics. It has also been a relief for low light conditions where facial recognition forced users to turn on the light, although this is still the preferred option in circumstances where there is a lot of noise. Complementing the two depending on the environment is the most effective solution.
Many have wondered about the difference between two a priori very similar concepts, voice recognition and voice biometrics. Although they share certain technologies, their purpose is very different.
While voice biometrics tries to identify one user to distinguish him from another, voice recognition is a tool for communication between people and digital devices. Also known as automatic speech recognition (ASR), it performs operations related to phonetics, syntax and semantics to understand what, through his voice, a human wants to express with an acoustic and sonorous message.
This computational tool is used to transcribe and recognize a message, as opposed to a tone. As we have seen, active voice biometrics makes use of voice recognition to understand credentials (passwords, phrases, numbers...) issued by an acoustic medium.
The two technologies can work together and function for both or different purposes by adding certain additional functionalities. The uses and applications of speech recognition per se range from automatic dictation to message interpretation.