BBC experiments with voiceprints
August 2, 2017
By Colin Mann
The BBC has worked with Microsoft to build an experimental version of BBC iPlayer that uses artificial intelligence to allow individuals to sign in to BBC services using their unique voiceprint and to talk to their TV to select what they want to watch.
Writing in the BBC Internet Blog, Cyrus Saihan, Head of Digital Partnerships, says that whether it is David Hasselhoff talking to his car in Knight Rider or Iron Man talking to his virtual assistant JARVIS, the idea of being able to talk to computers and for them to be able to understand who we are as individuals has been a science fiction fascination for decades.
“We may not be at JARVIS levels of artificial intelligence just yet, but artificial intelligence and voice interaction are fast-developing technologies that are already available to consumers. These technologies could also have interesting use-cases in TV so for our experiment, we wanted to explore how your TV – just by hearing the unique sound of your voice – could give people a more intuitive and more personal service in the future” he says.
Revealing details of the voiceprint experiment, Saihan says the ability of humans to communicate with each other by talking is one of our species’ most unique traits. “As the technology around us continues to evolve, it is interesting to consider how we might soon be talking naturally with the range of digital devices that have become such an important part of everyday life for many,” he notes.
“With voice controlled interfaces such as Amazon’s Alexa, Apple’s Siri, Google’s Assistant and Microsoft’s Cortana starting to gain popularity, there is a good chance that in some situations, speaking to a computer will be the main way that we interact with many of our digital devices,” he suggests.
In terms of a user talking to the TV, he says that just like the fingerprints of your hand, you have a voice that is totally unique to you. “In our experiment, by recognising the individual characteristics of your voice (tone, modulation, pitch etc), processing that information and then matching it to a sample of your voice stored in the cloud, artificial intelligence software checks that you are who you say you are and then signs you in, without you having to type anything,” he advises.
“Once the computer has been trained, the next time that you want to sign in, instead of having to type in your user name and password, you just have to say your name and a phrase. What we have built here is only a proof of concept and we are still at the very early stages for voice interfaces. Our experiment focussed on getting the basics right – creating a working internal prototype that allows you to sign in using your voiceprint. Once signed in, you can see all of the editorially-curated programmes and personalised recommendations that you normally would,” he says.
“As well as letting a user sign in to BBC services using their unique voice instead of a password, our internal prototype also gives a user the option to select what they want to watch by talking to their device. For example, saying ‘BBC…show me something funny’ brings up a selection of comedy programmes. If you say ‘BBC…what’s going on in the world?’ the BBC News channel turns on and starts playing. Saying ‘BBC… put EastEnders on for me’ starts playing the latest episode,” he reports.
As to what the future could hold, he noted that whether watching a football match or a quiz show, most of us have at some point shouted at our TV, perhaps half expecting it to hear us, know who we are and respond to us, and that in the future, we might find that it does!
“As the technology advances, voiceprints and artificial intelligence could enable even greater levels of personalisation. For example, if you’re watching a programme on your tablet on your way back from work then, later on, when you’re settling down on the sofa, your TV could ask you if you wanted to carry on from where you left off. You might respond ‘No thanks, is there anything new I might like?’ and be offered some suggestions,” he suggests.
“If we look further into the future, when artificial intelligence and machine learning have advanced sufficiently, you could end up in a conversation with your TV about what’s available to watch now, whether you like the sound of it or not, whether there’s something coming up that you’re interested in, and what you like to watch when you’re in a certain mood. All the time, your TV service would be learning about your preferences and getting smarter about what to suggest and when,” he adds.
“There could be interesting scenarios in a typical family setting too. Just by listening to the voices in the room, your TV could automatically detect when there are multiple people in the living room, and serve up a personalised mix of content relevant to all of you in the room. When your children leave the room to go to bed, BBC iPlayer might hear that the children are no longer there and then suggest a different selection of content for you and your partner. All of this personalisation could happen without anyone having to press a button, sign in and out or change user profiles,” he suggests.
“This was an internal experiment, designed to help us better understand how emerging technologies could impact the media industry and provide us with an opportunity to improve the experience for our audiences in the future. It’s an area that we are keeping a close eye on and adds to some other internal projects that we are working on,” he reveals.