In a bid for a larger slice of the voice recording market, Spescom DataVoice has set its sights on developing some cutting-edge speaker verification and emotion detection technology WHILE THE UTTERANCE "Your call is being recorded for quality assurance purposes" is hardly reassuring to Joe Public calling his insurance company to claim for his latest fender-bender, telephonic transaction records are crucial to the operational efficiency of said company, and many others like it, particularly in the insurance and financial services sector.

Spescom DataVoice, a subsidiary of the Spescom Group, caters to this very market with its cutting-edge recording technology, and to any other organisation engaging in transaction recording, including contact centres, trade floors, emergency services, utilities services and governments. Its solutions, which range from small single-box systems to large enterprise systems deployed on a national scale and capable of recording more than 120 000 calls an hour, are deployed in 40 countries around the world.

Spescom DataVoice has been developing communications recording solutions since 1989.

In fact, the very next year it pioneered the world`s first digital voice recorder.

In 1994, DataVoice produced a second-generation recorder capable of recording E1/T1 telecoms lines, and in 1997 it produced a third-generation product supporting up to 240 telephone extensions per platform. In 2000, the fourth-generation recorder was released to enable the recording of true digital inputs such as VOIP yet another world first, as well as agent workstation screen activity.

ARE YOU WHO YOU SAY YOU ARE?

CEO Viv Crone, who also doubles up as group chief technical officer, has now set his sights on perfecting the company`s speaker verification and emotion detection technology, in a bid for a larger slice of the voice recording market.

Identity theft is on the rise, a symptom of the increasing number of everyday transactions taking place online or over the phone through contact centres. Personal interaction is becoming more rare, and with it the ability to confirm someone`s identity by their physical presence.

An interesting statistic, as reported by CIFAS, the UK`s fraud prevention service, is that cases of false identity and impersonation fraud grew by 600% between 1999 and 2004. And the amount of money involved is staggering. CIFAS estimates that the 120 000 cases reported in 2004 alone cost the British economy 300 million.

"Clearly, existing tools for managing identity need to be improved by additional layers of security to increase the probability of a positive identification," stresses Crone. But, in a call centre this must not slow down the call-handling process, since these organisations are used precisely because they are cost-effective, convenient and efficient."

Security experts frequently talk about two- or three-factor authentication. We have all become familiar with passwords - `something you know` - as a method of protecting our virtual identity. But additional layers, like biometrics, add `something you are` to the security equation. The only biometric that works with a phone call is automatic voice recognition. A biometric measurement of a voice can be achieved by mathematically modelling the vocal tract to create a voice print, the vocal equivalent of a fingerprint, explains Crone.

Spescom DataVoice`s offering in this space is nearing completion. "We`re busy packaging it," notes Crone, who goes on to describe it as the best in the world. But it is early days yet for the technology, he admits. "There is more and more interest in this technology, particularly from the financial sector," he maintains.

HOW ARE YOU FEELING TODAY?

Another technology that is on the horizon is emotion detection software, which allows a provider to detect either a caller`s or call centre agent`s emotion on a call and analyse that emotion to respond appropriately. Admittedly, emotion is difficult to detect, whether by man or machine. Parameters that best define emotion in speech - be it anger, fear, happiness or sadness - are pitch, speech rate, intensity or volume, speech quality and word usage (like `unhappy`, as in "I`m unhappy with your service"). These parameters are relatively easy to measure, though not as easy as inferring emotional content.

The company`s team of 40 dedicated engineers and developers, based in Stellenbosch, have evolved speech analysis applications like Recall+, which allows a user to search through recorded calls for specific speakers or speech content, or a combination thereof. Finding problem calls is easy using its N-Best processing approach, whereby the system returns a number of results, sorting the data to provide the most likely result at the top of the list.

Gender and language identification are the challenges currently facing the development team.

"Technologically we`re as competent as anyone in the world," says Crone. The $10-million business spends up to 20% of its revenue on R&D. "To play the game, you have to spend a certain amount of money. If you play tennis, you need a tennis racket," he quips.

About 40% of this revenue is generated overseas, predominantly from its UK-based office, which services the European and Middle East market, and also from its US operation.

Sales are meeting their budget expectations, however, they have yet to experience the full impact of ongoing marketing and business development plans, he says.

Quite a few projects, most of which can`t be mentioned yet, are keeping his team on their toes. It`s busy implementing an emergency services solution, similar to the centralised voice recording system it installed to complement the City of Cape Town`s digital TETRA radio communications system, used by the city`s emergency and municipal services.