IMI Magazine

I recognise that voice!

A friend recently revealed that on his hour-long journey home from work he can get through 15-20 emails on his Blackberry. To make full use of his day he deliberately leaves messages to read and reply to. This would be fine if he was on the train, but he's not. He's driving. Aside from the risk of injury or worse in a serious crash, if it was proved he wasn't concentrating on the road he faces prison.

The problem is biggest among younger road users. The RAC Foundation says nearly half of British drivers aged 18-24 admitted to texting on the move. Research by the Transport Research Laboratory found drivers who use their mobile phone to send written messages - texts or emails - dramatically increase the risk of crashing. They're prone to 'drifting' on the road and reaction times deteriorate by 35 per cent. That's significantly worse, says the TRL, than those who drank alcohol to the legal limit or smoked cannabis, who were 12 and 21 per cent slower respectively.

"We have to be realistic about this. People text from their cars. The industry must work on solutions to minimise the risk to their safety," said Fatima Vital. She's marketing manager for Nuance Automotive Global, a leading supplier of voice recognition systems for cars. More than 25 million already feature Nuance technology, and it's in another 20 million aftermarket sat-navs.

Safe way forward
She believes talking to the car to perform tasks is the only safe way forward, adding: "We're currently working with car manufacturers so people can dictate and send their messages from their vehicle. The first deployment will be in the middle of next year." Fatima won't be drawn on which brands will get it, but Nuance software is in premium products such as BMW and Mercedes. It's also part of Fiat's Blue&Me and Ford's SYNC systems, both developed in conjunction with Microsoft. The latter will be launched in the UK in the spring, on the next generation Focus.

There are two issues to be addressed. Firstly, a message arrives and you want to reply to it; secondly, initiating and sending a new message. Nuance believes the ideal scenario is that you press a steering wheel button to interact with the car, just as you do now. But there's a new element with sending text - the validation process, or having the car read your proposed message back to you so you know it's right.

"The correction mechanism is the most important element to having a good user interface. We are working out what's a good way to engage with the driver for message validation. We are working with the car makers to develop it in a way that meets the needs and requirements of all parties," said Fatima.

'Visual deviation'
She acknowledges the overriding focus is safety of the driver and other road users. Nuance's own research shows that texting is 23 times more distracting than a face-to-face conversation with a front seat passenger. Fatima explained that 'visual deviation' - the posh name for taking your eyes off the road - is the issue. "To text while driving, you need to look at the screen frequently and people will typically do it while inputting a couple of letters at a time." She said visual deviation can be up to 40 per cent while texting, but with the technology in development, it's less than 10 per cent.

Speech recognition isn't as new as you might think. It dates back to the early 1950s, but the first real device was the IBM Shoebox, named for its size and shape. Ten small lamps were labelled from 0-9 and the corresponding light came on when the number was spoken into the connected microphone.

It worked because each word is made up of two or three distinct sounds. Each of these is high, middle or low-pitched and when passed through audio filters it's possible to work out what's being said. Today's technology is more hi-tech - it uses a mixture of acoustics, a dictionary and the rules of grammar - but operates in broadly the same way.

Voice recognition in cars started in 2004, controlling basic functions such as air-con. Next came music, when commands such as "play CD one, play track two" became the norm. Since then it's spread to sat-nav, mobile phones and - with the massive rise in external MP3 players - music.

Language recognition
Beyond texting, the next breakthrough is natural language recognition. It means you won't need to say: "Point of Interest search, new search, fuel stations, search nearby, search cheapest," in a rigidly structured format. "I need petrol, where's the cheapest?" or anything similar will do the job.

Fatima believes the next big thing will be 'connected services', which means logging on to the internet from behind the wheel to access information. "The complexity of the task means the only way to do this is by voice-activated searching," she explained.

Examples of how it could work sound appealing. You're driving into town for a meeting and need to park. Ahead of your arrival you can search for an empty bay and reserve it with a deposit on your credit card. As you leave, you pay the balance of what's owed.

You then head to the airport to pick up a colleague and check live flight data on the way to see if he will be on time. His plane is delayed so you stop, searching for your favourite restaurant chain's nearest outlet. A song you haven't heard in years pops into your head. You search an online database of music and live-stream it through the audio system. You could even buy it and save it to your MP3 player. And all this is done with your voice.

Such a world isn't as far away as you might thing. Audi's new A8 features Google searching, and the logical next step is that car makers will develop 'apps' for cars, so the homescreen of your vehicle's centre console screen will look like an iPhone.

"It's an issue that's under discussion at the moment," confirmed, Fatima, adding that future interfaces already in development by Nuance are being designed to cope with a future that looks like this.

It should make life easier - and much safer - for my friend and millions of other motorists. Or we could all just turn our phones off and enjoy the drive.

Richard Yarrow