Voice identification: opportunities and limitations
We were talking about automatic speech recognition as a support for home automation and pedagogical support. Today, voice in the development and use of new technologies is essential. For all sectors, orality is making a strong comeback. Voice recognition is very useful for several functions. Therefore, here we focus on voice identification. It is more than « simple » recognition, voice identification allows authentication in different contexts.
Voice recognition and voice identification
To begin, let’s briefly review what speech recognition is. According to futura-sciences.com, it is the « Recognition, by a computer system, of the elements or words of a voice message ».
We must be careful, because automatic speech recognition – or voice recognition – is different from identification. Speech identification is linked to the identity of the individual using certain technologies. It is a matter of authenticating the person, then authorising him or her to access certain data, make certain payments, etc.
Voice identification is therefore closely linked to biometrics. It is a priori less intrusive.
Voice identification: opportunities and limitations
Voice identification is a lever to take with a pinch of salt! It is synonymous with opportunities, but you have to be careful to choose the right tools and to ensure the security of your data.
According to techniques-ingenieur.fr, Google « is experimenting with payment by voice identification through the microphone of the Google Home connected speakers. It is no longer necessary to grab your smartphone to validate a transaction on the Play Store using a password for example: the voice of the authorized user is enough« .
So authorizing payments by voice is not entirely new, as some banks already offer it to their customers. But by taking the plunge into voice identification, Google is really rolling it out on a large scale.
In fact, there are several advantages and opportunities to voice identification. This biometric technique makes certain purchasing acts more accessible and, above all, it’s faster.
Moreover, if banks opt for voice identification systems it is not by chance. Biometric identification responds to security threats, various frauds linked to identity theft and even cybercrime in general. However, although the reliability and speed of identification and authentication is increasing, this technology also has certain limitations.
Security, banking and voice identification
According to meilleurbanque.com, biometric authentication consists of « automatically verifying identity or identification. This verification is based on individual physiology or behavioural characteristics. Today, banks are testing a voice identification system« .
This whole process is precise, complex. For banks, security is central. And after physiology and behavioural analysis, voice is a source of future, progress for authentication.
How voice identification works
Specifically, when you speak to the Google connected speaker, the system adjusts. It takes into account the particularities of your voice. It’s your tone, your voice variations that come into play for voice identification.
These singularities are extracted from an audio signal lasting about ten seconds – at best 30 – before a vocal « signature » is deduced. The system compares different voices and superimposes the signatures. It is all this expertise that makes it possible to validate an access, a transaction, a successful or unsuccessful authentication.
In addition, such systems – also called i-vector – can eliminate ambient noise and above all standardise the sound variability according to your smartphone. This is because they are familiar with the specific nuisances of smartphone models. Thanks to this development, the error rate is considerably reduced.
Framed voice identification
Voice identification is highly regulated in France. Banking institutions and the companies that use them are closely monitored.
According to alain-bensoussan.com, the CNIL has « authorised the implementation of customer authentication devices using voice recognition. These authorisations were granted on a trial basis to banking institutions that wanted to strengthen the authentication of their customers in two situations: when the customer makes an online payment and when the customer logs on to his customer account online« .
The main objective is to resolve password issues. For example in case of lost or forgotten passwords. Voice identification is a means of authentication that is accessible under any circumstances.
Today there are two key situations where the user will rely on voice identification:
The user wishes to make a payment on an e-commerce website. In this case, he receives an automatic call and must then say an authentication sentence. This automatically triggers a pre-filled payment form – with the account number, validity date, cryptogram, etc. – which is then sent to the user.
The customer user wants to access his personal bank account. He receives an activation code and says a passphrase there as well, which enables him to authenticate himself.
Voice identification is accessible to everyone without distinction. The only question that can be asked at this level is the consequences of a voice that does not recognize itself, which no longer allows the user to authenticate himself. If the user has a cold, with a congested voice, will he be blocked in his banking and administrative transactions?
Identification by voice
Finally, it is important to note that voice identification – or identification by voice – falls within the scope of GSPD. As biometric data, it is monitored. However, for the time being, voice is not recognized as sensitive data. This is surprising because the voice reflects a person’s condition, origins, age, gender… Voice identification and its protection will certainly continue to evolve.
Whether for banking transactions, real estate transactions, medicine, home automation, education… Voice identification is a complex process that has many possible applications but is still in its infancy. Biometrics is not yet widespread as in some futuristic films, but the advantages and accessibility by voice are readily recognized.
We hope that this article will have helped you to see more clearly the differences between automatic speech recognition, voice identification… And above all the strengths and weaknesses of the process. Do not hesitate to contact us now to discuss your problems related to automatic speech recognition!
Authôt. You speak. We write.