AI in Russia Uses Voice to Flag Risk of Chronic Disease
New app analyzes short voice recordings to assess the likelihood of cardiometabolic disorders.
Graduates of the Digital Department at Sechenov University have developed an AI-based application that evaluates the risk of cardiovascular and metabolic diseases using a person’s voice.
The app is designed to be simple: users record a short audio clip on their smartphone and receive results within seconds. The system automatically generates recording instructions, analyzes vocal features, and estimates the probability of developing health conditions. Developers adapted the model to the specifics of Russian-language speech.
Speech as a Health Marker
According to Denis Kuznetsov, a graduate of Sechenov University’s Digital Department and a PhD in biological sciences, speech involves the respiratory system, laryngeal muscles, articulation mechanisms, and neural regulation processes. As a result, changes in the body’s condition can alter voice parameters. Such variations are nearly impossible to detect by ear, but digital audio processing and machine-learning methods can identify them.
Earlier, IT Russia reported on researchers in Novosibirsk who learned to assess depression using human speech. Their prototype digital service provides a full data-processing cycle, and scientists trained a neural network model to classify voice samples across four levels of depression severity.
