The Covid-19 epidemic has stimulated research around the world for ways to early detection of infection. Since the virus leads to a variety of symptoms that change from person to person the quest for a symptom based detection is difficult. On the other hand, even the possibility to raise a red flag for the probability of infection would be a significant step forward in monitoring the epidemic and fighting it.
From the very beginning a number of researchers focussed on the alteration of voice in people affected, alteration that can become detectable even before the person notices. Also, coughing is often a manifestation of the virus and the way of coughing can be an indication of infection. Hence researchers have developed a number of apps that once installed on the smartphone can use its microphone and signal processing to analyses changes in your voice patterns. This is done first by creating a digital signature of your voice, something that can take a week (it depends also on how often you use your phone for talking). The analyses of the voice, signal processing, continues and any divergence from the acquired digital signature (your normal way of talking) is processed by an artificial intelligence based software to evaluate the risk of having Covid-19.
Here comes the problem: the AI software needs to learn what “divergence” can point to an infection and this is done using a huge amount of data derived from many (the more the better) people. These data should report both the type of divergence and the information of that person being or not being infected. These data are difficult to harvest and, obviously, are sensitive, personal, data.
A recent study indicates that the technology is not a problem: our signal processing technology and the smartphones are suitable to detect any subtle nuances of our voice. The problem lies in the difficulty in getting sufficient volume of “reliable” data (be certain that a given set correspond to infected persons, whilst the other to not infected person).
The study makes a call towards institutions and governments to agree on standards and open data framework so that the potentially huge amount of data captured by billions of smartphones can be put to use.
It is interesting to note that technology has reached a point where limitations are in the context in which it is used, not in its capabilities.