Decoding a computer-generated voice

26 March 2006


Anna - Welcome to the King's School in Ely where this week we're going to be astounding you with sound and finding out all the glorious glorious things about the science behind music. And here to help us do that this week, we've got our amazing science guru Wendy. So , can you introduce yourself?

Wendy - I'm Wendy Sadler and I'm from Science Made Simple and we take loads of science stuff around to schools and try and make it fun and really exciting.

Anna - What is it that we're exactly going to be doing today?

Wendy - Today we're going to be testing out people's listening skills and we're going to see how good they are at spotting some words within an electronic voice.

Anna - Wow that sounds pretty intriguing. Also here with us, we've got some student helpers. So can you tell us your name and your ages please?

Emily - Hi I'm Emily and I'm ten years old.

Anna - And what about you sir?

Matt - I'm Matthew and I'm nine years old.

Anna - And so Matt, what's your favourite thing about science?

Matt - It's probably the experiments.

Anna - And obviously most importantly, do you know anything about the science behind music Emily?

Emily - No!

Anna - That is a good job because that's exactly what we're here to tell you today. So what exactly is it we're going to be doing Wendy?

Wendy - We're going to be doing a fun experiment that's about the science of sound and also bout how our brains work. We're going to play a clip in a moment which is a very early attempt at a computer trying to speak. So it's a computer synthesising a human voice. The human voice is a very complicated sound. It's quite hard to work out what it's saying, so this is the challenge for everyone, to listen and see if they can work out what the voice is saying.

Anna - So people at home, you should be listening very carefully to what Wendy's going to play you here. You need absolutely no equipment whatsoever so there are no excuses this week. So Wendy's going to play this song for us now and I want you guys, Matt and Emily, to try and tell us what you think it says…. Any ideas?

Emily - No!

Anna - And what about you Matt?

Matt - I haven't got a clue.

Anna - Can we hear it again Wendy, and this time we'll listen a bit more carefully…. Were there even any words that you could pick out there Emily?

Emily - Absolutely none.

Matt - I think I heard sound at one point.

Anna - Ok so we've got a bit of a hypothesis there. Well we're not going to tell you exactly what it says quite yet because we want you people at home to have a think about what it said. We'll be playing it again later on in the show and we want you to phone in and tell us what you think they said. Thanks to Wendy, Matt and Emily and we'll be going straight back to the studio now.

LATER ....

Chris - Anna, let the cat out of the bag. What does it say?

Anna - Hello and welcome back to the King's School in Ely where we're here with Wendy from Science Made Simple and our student helpers Matt and Emily. Now what we've been doing is listening to some very strange computer generated voices and having no idea whatsoever about what it's saying. So what are we going to do now then Wendy?

Wendy - Well before we reveal the answers, we'll play it one last time to give people their last chance to see if they can get any words. I think Matt picked one up before.

Anna - Ok well that's still all Greek to me, so Matt can you reveal exactly what it says here?

Matt - Sound and music can be used for the synthesis of speech.

Wendy - It sounded a bit different when Matt said it!

Anna - It did indeed sound different. So can we actually see whether it actually said sound and music can be used for the synthesis of speech?

Wendy - Ok, this is exactly the same track again and I haven't changed it at all.

Anna - Emily, could you hear that this time?

Emily - Yes.

Anna - Are you sure?

Emily - Well I can hear bits of it now that I know what it's saying.

Anna - Shall we listen to it one more time? Sound and music can be used for the synthesis of speech.

Wendy - Matt actually picked up one of the words because he got sound right.

Anna - Wow that's absolutely amazing. I really can hear that. How did you make that sound in the first place using a computer voice? How does that all work?

Wendy - Well this is quite an early attempt at computer voice synthesis, but it's actually quite hard to copy a human voice. Every time you speak, you've got two elements of sounds going on. One is the pitch part, so you're essentially singing as you speak. You can speak really high or you can speak really low. That's one part the computer has to generate. The second thing it has to do is generate these sounds that are like percussion. They're called fricatives which are things like 'p' and 'k' and they're very hard sounds to form. Now one of the problems with this voice is that when you hear it you'll hear it's all on one pitch. This makes it sound very robotic. You immediately know that it's not human being and it's very hard to copy a human voice. It's almost as unique as your fingerprint. The reason it's very hard to understand is that you've only got one pitch and it's not very good at making those 'p' and 'k' sort of sounds.

Anna - But why is it that as soon as you showed us exactly what it was and showed us those words sound and music can be used for the synthesis of speech, that then all of a sudden we could hear it?

Wendy - Well your brain is really powerful at filling in the gaps and obviously if you suggest to someone what they're meant to be hearing then when they hear the words in their head and hear the voice again, they can put in all the stuff that was missing.

Anna - So is that kind of like when I'm listening to someone in French and I only know a few French words, I can fill in the gaps and kind of get what they're saying?

Wendy - Yes. You can put the sentence together effectively by the words that you're missing. In fact it's a little bit like when people suggest that they can hear hidden lyrics in songs played backwards. If someone suggests to you that you can hear these certain words, then you're likely to hear them because you are expecting them.

Anna - And Matt, you had a question.

Matt - It seemed really amazing that when we knew what it was saying, that it seemed so much clearer.

Anna - Yeah that really was amazing and now do you feel that you understand exactly why that was, with our brains filling in the gaps?

Matt - Yeah.

Anna - And what did you think of the experiment then Emily?

Emily - Quite weird!

Anna - And were you expecting it to actually say anything in the end?

Emily - No.

Anna - Well I have to say thanks very much Wendy for that. Computer voices and our brains being amazing and filling in the gaps. Well that's it for this week and that's enough from the King's School in Ely. Thanks very much Emily, Matt and Wendy. Did you enjoy yourself Wendy?

Wendy - Yeah it was great fun.

Anna - Thanks very much. We'll be back doing some more kitchen science next week somewhere in the Eastern region. So goodbye for now.


Add a comment