HAVE A NATTER

Acoustic engineer TREVOR COX (pictured below) makes everything from concert halls to washing machines sound better, and now he’s turning his attention to the mysteries of human conversation. HELEN GLENNY asks him what he’s learned…

When do we start learning speech?

From about the third trimester of pregnancy a baby can start hearing the vague sounds of a mother’s voice, and when a baby is born it recognises that voice in preference to any others. It doesn’t recognise its father’s, because it hasn’t had that same intimate connection.

How does the voice change over a lifetime?

There is a big change at puberty – that’s when we get our adult voice. It’s most obvious with boys, but female voices drop too. But what’s remarkable about the voice is that it doesn’t really deteriorate with age until you’re very old; I’m in my 50s and I’m covered in wrinkles and my body is gradually falling apart, but my voice isn’t that different from when I was 20.

You say we make assumptions about people based on their voices…

Yes, and one that I’m fascinated by is the gay male voice. We have this situation where people assume that gay males automatically talk at a higher pitch, but if you compare straight and gay men, and do measurements, there’s no difference. During my childhood, the only gay males I knew of were camp characters like Larry Grayson. We didn’t have gay male role models who spoke as everyone else did, so I expect that stereotype might wane over time – hopefully it does, because it’s incorrect.

Speech isn’t purely the domain of humans any more. Google’s voice assistant can book appointments over the phone, and it sounds real. How?

Those voices are already pretty good, but Google added the ‘uhhuhs’ and ‘mmms’ that we use when we’re thinking, which makes it sound more natural. Another clever thing is that they did it down a phone line, so if there’s any roboticism, the listener will assume it’s a mobile

phone issue.

As well as talking, robots are singing now too, right?

The best example is Hatsune Miku who does Japanese pop. People turn up to Hatsune’s concerts and listen to her sing, even though she’s just a synthesiser and light projections. There’s a robotic aesthetic in Japanese pop, so you’d struggle to pick that her voice is synthesised. There’s a lot about the voice of a singer-songwriter like Adele that we just can’t do with synthesis yet.

What about robot actors?

I went and saw a robot perform ‘Alas, poor Yorick!’ [a speech from Hamlet] recently. It was funny, because the speech is all about mortality, something a robot doesn’t suffer from – I guess until a software update means it doesn’t work any more.