New innovation explores how smartphones could help detect disease
Our voices may reveal more about our health than we realise. Subtle changes in the way we speak can be linked to disease, sometimes years before a diagnosis is made. Coen Elemans now wants to develop technology that could make it possible to capture this information using nothing more than an ordinary smartphone.
Many of us have regular health checks. We have our blood pressure taken, undergo ECGs, step onto the scales and give blood samples. In the future, we might be asked to do something else as well: record our voice on our phone and send the recording to our doctor.
That is because our voices can carry information about our health that we cannot hear ourselves. Scientists have only relatively recently begun to explore this potential, but the field is developing rapidly.
Research suggests, for example, that Parkinson’s disease can cause subtle changes in speech and voice that the human ear cannot detect, but which can be measured up to ten years before a conventional diagnosis. This has fuelled growing scientific interest in the possibility of using voice as an early biomarker of disease.
Almost all have a smartphone
“Nearly all of us have a smartphone, so recording our voice and sending it to our doctor would be incredibly easy for us and inexpensive for society. It could provide an entirely new layer of information that might help us in ways we have never had access to before,” says Coen Elemans, Professor of Bioacoustics at the Department of Biology, where his research includes how animals and humans produce and use their voices.
There is, however, a major obstacle to overcome before voice-analysis technology can make reliable use of smartphone recordings. Coen Elemans now wants to tackle that problem.
He has received DKK 1.2 million from the Faculty’s innovation fund for the project SMART-VOICE – Unlocking Smartphones for Scalable AI-Ready Voice Science.
The voice’s higher frequencies are important
Recording your voice on a phone is easy enough. The voice is usually perfectly clear, and most of us would probably agree that the recording sounds just like the person who made it. But behind the scenes, a lot is happening that most phone users never have reason to think about. Rather than simply capturing the sound of our voice, the phone processes the recording in numerous ways to make the voice sound as clear and intelligible as possible.
“By the time you listen back to the recording, the phone has already processed it automatically. What you hear is no longer a raw recording, and that becomes a problem if you want to analyse it for diagnostic purposes,” says Coen Elemans.
Making matters more complicated, researchers do not know exactly how individual phones process raw recordings, and manufacturers do not provide this information. As a result, researchers cannot simply feed a smartphone recording into an analysis tool and expect reliable results. Phones often remove higher frequencies from the voice, for example, which is a serious problem if you want to study the physiology of the vocal folds.
Could become hugely valuable to our healthcare system
To tackle this, Coen Elemans and his colleagues will systematically investigate how different phones and recording apps alter the audio signal.
Among other things, they aim to develop standards, correction factors and measures of uncertainty that will allow researchers to determine whether, and under what circumstances, smartphone recordings can be used for voice analysis.
“If, in the long term, we can develop technology that can reliably isolate raw voice recordings and send them to an analysis platform, it could be hugely valuable to healthcare. We already know that voice analysis can be used to help diagnose disease,” says Coen Elemans.
Cognitive disorders and depression
Parkinson’s disease is among the conditions most extensively studied using diagnostic voice analysis. A scientific review published in BMJ Evidence-Based Medicine in 2026 examined 96 studies of voice- and speech-based biomarkers across 20 diseases.
Cognitive disorders, depression and Parkinson’s disease were among the most frequently studied conditions. Overall, the models correctly identified 80 per cent of people with mental health disorders. The figure was 85 per cent for cognitive disorders and 81 per cent for movement disorders, particularly Parkinson’s disease.
The project SMART-VOICE builds on Coen Elemans’ research into voice acoustics as well as research into voice analysis by Oliver Niebuhr, Associate Professor at the SDU Centre for Industrial Electronics.
Natural science for societal needs
Innovation is one of the many ways in which research can create value for society. At the Faculty of Science, fundamental research forms the foundation of our work and accounts for the vast majority of the research we conduct. However, in some research environments, fundamental research and societal value creation go hand in hand and reinforce one another. This is why we actively support innovation wherever new knowledge has the potential to generate solutions, partnerships, companies or other forms of societal impact.
Meet the researcher
Coen Elemans is Professor of Bioacoustics and Head of Research at the Department of Biology. His research focuses on how animals, including humans, produce and control sounds for communication.