When the body develops a cold, changes in speech occur. A new non-invasive method can detect impending colds early based on speech.
In a research project involving experts from the Rheinische Fachhochschule Cologne (RFH), a classification model based on deep learning was examined, which can use language to predict whether a person will develop a cold or not. The associated study is published in the journal “Circuits, Systems, and Signal Processing” (CSSP).
Typical symptoms of colds
A cold is a viral infectious disease of the upper respiratory tract. Those affected can develop various symptoms such as cough, sore throat, runny nose, sneezing and hoarseness, the researchers explain.
Because the vocal tract is also affected, a person’s speech changes when they develop a cold. For example, according to the team, the speech of a person with a cold is characterized by a lower pitch.
With the help of the new method, so-called general cold information can be recorded from speech signals. Based on so-called deep learning, the researchers trained a model that can categorize signs of colds in speech.
Predict colds from afar
This model is less complex and uses a smaller set of features, but produces comparable results to other modern diagnostic methods, according to the team. This enables automatic identification of colds and related illnesses.
Thanks to the new technology, it is no longer even necessary for the person with a cold to go to a practice to be examined. The new method can also be used remotely as part of so-called telemedicine.
If there are early indications, some home remedies could also be used against the common cold, for example nasal rinses with a saline solution to clear the nasal cavity of germs and accumulated mucus and can contribute to both the prevention and treatment of acute colds. (as)