Russian Scientists Go Ahead With Voice-Based Depression Diagnosis

Researchers at Novosibirsk State University have set up a digital tool for diagnosing depression, the university’s press service said. Faculty members and students from NSU’s Institute of Medicine and Psychology proposed an automated approach based on analyzing a person’s vocal intonation. Speech is treated as a natural biomarker of mental state, with changes in vocal energy offering clues about a speaker’s mood and condition.

Trained on Interviews

The research team trained a neural network on 90 interviews with different participants, including people diagnosed with depression. Using the wav2vec architecture, the scientists extracted vector-based acoustic features from voice recordings.

The tool is expected to help clinicians make diagnoses faster and more accurately. According to the researchers, its potential use is not limited to depression. With additional data, the model could be trained to detect other mental health disorders based on voice patterns.