Amazing: New Israeli Technology Detects Diseases Through Voice

Did you know that through the sounds we make from our throat, we can derive various data about a person - their character, temperament, and even what diseases they may have? Dr. Yossi Kashet, a voice analysis expert from Bar-Ilan University, has been leading in recent years along with his team members - a groundbreaking system through which countless surprising details about a person can be discovered, just through the sounds coming from their throat.

Years of development have yielded impressive results, although "when I started working in the field of speech recognition, I realized there was some missed opportunity or issue in its use," explains Dr. Kashet in an interview with 'Haaretz'. "It is a field that quickly transitioned to industry without sufficient academic research. Its use is still limited, and it doesn't always work as it should."

Nevertheless, he manages to achieve, through this method, an accurate diagnosis of various diseases in the vocal cords, as well as other information he can extract about the 'speaker' - such as answering whether any distress is heard in their voice that needs immediate attention, or if it is a 'fake' distress.

Today, Dr. Kashet's Lab for the Study of Human Vocal Cords is considered one of the most important labs, having gained an international reputation in the field. According to Kashet, there are many parameters in speech that, firstly, are beyond our understanding, and secondly, are uncontrollable. "Like the length of the consonant," he explains. "When I want to communicate with someone, I adjust the length of my consonant to that of my partner. They do the same, and actually, our consonant lengths converge.

"If I produce the consonants in a long manner, and my conversation partner produces them in a short manner - then in our conversation, each will adjust their consonant length to the other. This phenomenon can be exploited to measure the potential quality of the relationship between the speakers." Dr. Kashet noted another example, which is our tendency to match our accent and speech style to whom we are talking with.

So what is voice analysis, and how is it conducted in practice?

"To extract the hidden information in the human voice, a musical ear is not required," explains Dr. Kashet. "What is needed are machines with powerful computational capabilities, large databases of vocal information as raw material, the ability to ask the right questions, the ability to identify unique characteristics, and the development of algorithms that provide the answers."

In the past, such information was collected manually by researchers. For instance, in the field of linguistics - basic research in acoustic linguistics, for example, "required researchers about 3,000 hours of research on average, with most of the time spent on the manual marking of the acoustic signal. Today we work on a data amount that is ten times everything existing in the professional literature, and it takes a few minutes."