Speaker
Verification, (sometimes called Voice Verification), has
been in the news recently here in the UK with the first major initiative
being commissioned by BT following a competitive tender. The winning
supplier is the Israeli company, and ICR partner, ‘Persay’.
Many blue chips have been looking at the technology for a number
of years, indeed ICR was involved in a free-speech speaker verification
trial with Persay, CCIR and Lloyds TSB 3 years ago. But BT has been
the first to take the plunge in the UK, and the implementation is
likely to be watched very carefully by other organisations. BT is
to incorporate Persay’s technology within its URU (You are
You) Identification and Verification process to reduce fraud.
Put the potential scope of this service in context, false identification
(people claiming to be someone else) leading to fraud, costs England
alone an estimated £1.3 billion per year.
So, for those of you who are not familiar with Speaker Verification
- its purposes and advantages - this issue of Talking Technologies
provides an overview of the technology and its application.
What is Speaker Verification ?
The basic principles behind Speaker Verification are to identify
and verify callers over the phone and to prevent fraud by means
of a ‘voiceprint’.
There are a number of competing systems, but they generally fall
into three categories :
Free speech – Registration and subsequent
verification of a voiceprint takes place in the background, during
a conversation between the caller and an agent
Random Digits – the caller is asked
to speak a set of random digits, generated and then checked by an
automated system. |
 |
Vocal
Password – the caller registers a password (usually
a memorable phrase) and is prompted to repeat it when calling
the organisation subsequently.
The most popular forms of Speaker Verification deployed to date
are the use of Random Digits and Vocal Password.
Benefits of Speaker Verification
A number of benefits exist for both the organisation and the caller
which are as follows:
- Fraud can be significantly reduced - many organisations currently
depend on an Identification and Verification (ID&V) process
to control access to a wide variety of transactions and information
services, either via an automated system or a call centre agent.
The ID&V process is based on information which is known
only to the caller such as account numbers, PINs, passwords,
date of birth and so on. However, "Identity Theft"
is a rapidly growing crime, and if a fraudster is able to gain
access to this information they have instant access to a person’s
accounts. Speaker Verification prevents this: access is only
permitted when the correct information is spoken by the correct
person.
- If an organisation wishes, the process of ID&V can almost
always be conducted by an automated process rather than via
a human resulting in reduced costs for the Contact Centre.
- The ID&V process can be significantly streamlined with
reduced call time, to the benefit of all concerned.
- The caller is likely to be delighted that security has been
improved and that any risk of access to their information and
private transactions has been denied to others.
|
|
Registering
and Storing Voiceprints
Use of Speaker Verification initially requires the storage of a
customer’s voiceprint. This is best carried out at the point
when a customer first registers for voice based services, and should
be linked with a secure verification process, to eliminate fraud.
The precise nature of the process deployed is dependent on the type
of verification system selected, and whether it is a new or existing
customer. ICR can supply model processes for each of the major types
of system.
With the exception of Free Speech systems, processes can be either
automated or handled by an agent.
Most commercial systems do not store a copy of the Voiceprint, but
only key characteristics extracted from it at the time of registration.
This not only ensures that demands on storage capacity are minimised,
but also that voices cannot be reverse-engineered from voiceprints,
providing additional reassurance regarding security. The latter
means that any customer fears about the storage of personal physical
characteristics are unfounded.
Conclusions
In our opinion Speaker Verification will become a valuable addition
to the call handling process for many organisations in the future,
reducing fraud, reducing call handling costs and benefiting the
customer.
Further Information
For further information on Speaker Verification you can download
ICR’s ‘Guide to Speaker Verification’ by
clicking here
and you can visit Persay’s website at www.Persay.com |
