“Bad news I’m afraid”, is one of the more considered approaches to calling the office to tell the boss you are sick, and won’t be coming into work today. I have no idea what the percentage of such calls are a ruse, simply an excuse for not going to work, or just a flat-out lie. However, whatever the motivation, the validity of those calls is about to verified by AI; in other words, AI just got a lot more dangerous and intrusive.
A research team at the Sardar Vallabhbhai National Institute of Technology in Surat, India, have found a way to decode the difference between a “cold” voice and a “healthy” one, regardless of how well you might try and hide the real situation. That frightening thought is even more frightening if you begin to think of other possible applications of this technology.
We all know that our voices change when we have a cold. We may not know that the common cold virus tends to inflame the vocal cords, which changes their acoustic properties. The tissue temporarily swells and, therefore, vibrates at a lower pitch, which tends to make people with a cold develop a deeper voice. What the researchers in India have done is to develop a technology that can determine if your deeper voice is real, or faked.
Human speech, like any musical instrument, does not produce single frequencies of sounds. Even the best-trained singers cannot hit pure notes like those that come from a tuning fork. The dominant notes in the human voice are, instead, accompanied by a series of higher-pitched overtones.
Together, these sets of notes fit into mathematical patterns called harmonics, with overtones having frequencies that are multiples of the original note. For example, the pitch of the second harmonic note is twice the frequency of the main note, and so on. The amplitude (loudness) of these harmonics in speech tends to diminish as they proceed up the frequency scale. The researchers at Sardar Vallabhbhai National Institute of Technology reasoned that the infection that occurs with a cold might alter how this process works. That’s bad news, I afraid, for all those “skivers” wishing to sneak a day off work.
To find out how this might work, the scientists made use of an unusual resource; recordings of 630 people in Germany, 111 of whom were suffering from a cold. Each person was asked to count out loud from 1 to 40, and then describe what they did last weekend. They were also asked to read aloud from Aesop’s Fable, “The North Wind and the Sun”, which has been a popular text for phonetics research since 1949.
By breaking down each person’s speech into its spectrum of component wavelengths, the researchers could identify the dominant frequency, and the associated harmonics, in each case. They then used machine-learning algorithms (AI) to analyze the relationships between the amplitudes of these harmonics and patterns that could distinguish the “cold” voices from “healthy” ones. Have I lost you yet?”
Scientists are now looking into other applications of this analytical tool; for example, Parkinson’s disease and depression, as well as head and neck cancers, all can affect a person’s voice. There are even research projects that propose to use the way we talk, write and walk as health analysis tools. I don’t know about you, but that, potentially, sounds as scary as hell!
The Indian team’s research is not foolproof, yet. The results have shown a 70% correct diagnosis rate. However, do you really want to take that risk when you call in with a fake cold?
I guess the only solution for the “faked cold” phone call problem is to shut up, but that’s rather a contradiction in terms, unfortunately.