Caller identification with voice biometrics
In the Contact Center, a voiceprint enables the caller to be identified and authenticated. Voice biometrics saves time at work – and customers no longer need to answer any security questions.
Text: Urs Binder, Images: © Alamy,
Sylvie M. has a question about her new fixed-network subscription and calls the Swisscom hotline. In the Contact Center, Markus H. takes the call and Sylvie tells him about her issue. Markus is able to give her the right information immediately, because he already knows which customer this is. Even if confidential matters are involved, such as payments, Sylvie does not have to answer any security questions or enter any PIN numbers.
This is because – in the background – a voice biometrics system has been analysing Sylvie’s voice characteristics from the beginning of the call and comparing them with the voiceprint that was saved from a previous call. Within a few seconds the system was able to identify the customer as Sylvie, and Markus was given the green light for the rest of the conversation.
Voice biometrics is the only biometric procedure that also works remotely: For other methods such as fingerprinting, iris recognition or even DNA comparison, there needs to be direct contact with the person. Speaker recognition with voice biometrics is a very reliable and efficient method of identifying persons based on individual characteristics of their voices – thus ensuring that a caller really is the person that they are purporting to be.
During the first call the system creates a “voiceprint” that records all of the identifying characteristics of the voice. Around 40 seconds of conversation are required for this. This voice print is then ready for identification in future phone calls. The system requires 10 to 15 seconds to identify the caller. There is no need for the caller to say a predefined text; they can communicate freely as the identification is performed in the background without the caller being aware of it. And the method very seldom fails – for instance, the voice is even identified if the speaker has a cold.
The method can also be used for sensitive areas such as the health sector and finance, because the voice can continue to be analysed during the call, enabling fraud to be detected, e.g. a switch of speakers during the phone call. Voice biometrics is also suitable for providing an additional security level in cases where particularly confidential matters are being discussed.
The voiceprint is not an audio recording and contains no personal data, as only meta data is saved. The connection between a person and a voiceprint is not obvious. Nevertheless, the customers involved should be able to give their consent to voice recognition, ideally by means of explicit consent via an opt-in or via a clearly communicated opt-out option. In the EU, an opt-in version is already prescribed.
Identification via voice biometrics enables a significant reduction in the length of phone calls for customer enquiries. Typically, identification and verification takes up around 30% of the total call time, and with voice biometrics this is performed completely in the background. For example, a large call centre with 3.8 million phone calls per year and an average call duration of 300 seconds saves up to 4,200 working days by using voice biometrics.
If additional voice technologies are used – see box – the processing time can be reduced by half. The benefits for callers are obvious: They no longer have to deal with annoying security questions or remember PIN numbers, passwords, customer numbers or contract numbers – plus they are serviced more quickly. Voice biometrics also supports fraud detection and enables unwanted callers to be blacklisted.
A voice biometrics solution is suitable for any scenario in which a contact / call centre is in use, and is worthwhile even if there are only a few hundred contacts per month. Swisscom has been using voice biometrics for several months for its own hotline and the experience has been very positive.
Swisscom offers the solution to interested customers for their own call centre. This always starts off with a proof-of-concept. This can be operated independently of the existing call centre infrastructure to enable the customer to test the services. These include not only caller identification via voice biometrics but also functions such as speech-to-text and AVR – and the Swisscom solution even understands Swiss German, in contrast to comparable products.
Voice biometrics is about recognising the individual characteristics of a person’s voice. Taken together they provide a voiceprint with which the person can be uniquely identified, validated and authenticated – completely independently of the content of their speech.
In this way, voice biometrics enables speaker recognition. Voice recognition is something different: Here the system identifies the language that is being spoken, while the content of the speech is analysed and converted into text. Voice recognition is used by applications such as IVR (Interactive Voice Response), AVR (Automatic Voice Recognition) and Speech-to-Text.
More on the topic