1
Biology
Respiration, phonation, motor control and cognition contribute to speech.
Voice biomarker guide
What are voice biomarkers?
Voice biomarkers are measurable characteristics of speech that may reflect aspects of a person’s physical, neurological, cognitive or psychological state.
The short answer
A voice biomarker is a measurable feature, or combination of features, derived from voice or speech and associated with a biological or health-related process. Examples include pitch variation, speaking rate, pauses, vocal stability and spectral characteristics. Artificial intelligence can analyze patterns across many such measurements, but the result must be validated for its intended population and use.
Why can the voice carry health information?
Speaking is a coordinated biological process. The lungs supply airflow, the vocal folds create sound, the vocal tract shapes it, and the brain coordinates language, timing and movement. Changes in respiratory function, motor control, cognition, mood or fatigue can therefore influence the resulting signal.
2
Acoustics
A recording captures variations in pitch, timing, energy and voice quality.
3
Analysis
Signal processing and machine-learning models identify relevant combinations of features.
4
Insight
A validated model produces a score, trend or screening signal for a defined purpose.
What does voice analysis measure?
Feature family
Examples
What it describes
Prosody
Pitch, intonation, rhythm
How speech rises, falls and varies over time.
Timing
Speaking rate, pauses, response latency
The pace and temporal organization of speech.
Voice quality
Stability, breathiness, perturbation
Characteristics of vocal-fold vibration and airflow.
Spectral characteristics
Energy distribution and frequency-domain measures
The acoustic composition of the recorded signal.
Linguistic features
Vocabulary, syntax and semantic content
What is said. Some systems use these features; others focus on how someone speaks.
Virtuosis primarily analyzes acoustic characteristics rather than the meaning of the words, which supports use across languages. Recording context and speech fluency can still affect a model and must be considered during validation.
Where are voice biomarkers being studied?
Mental health and wellbeing
Research has examined associations between depression and measures such as speech rate, pitch variability, prosody and voice quality. Reviews find promising results but also substantial variation in datasets, recording tasks and validation methods.
Neurological and cognitive health
Changes in articulation, prosody, fluency and language can be investigated in conditions affecting movement or cognition, including Parkinson’s disease and mild neurocognitive disorders.
Respiratory health
Voice, cough and breathing sounds can contain information related to respiratory function. Studies include asthma, infection and symptom monitoring, although performance depends strongly on the task and population.
Fatigue and communication
Timing, energy, vocal effort and variability may help characterize fatigue, stress and communication patterns. These applications require clear definitions and validation against appropriate reference measures.
What published studies illustrate
Research examples show how different health processes may influence different parts of speech. They also show why results must be interpreted in the context of the dataset, recording task and validation method.
Mental health
Depression
Studies have examined changes in spectral energy, pitch variability, speaking rate, vocal control and pauses. In one small preliminary study, an acoustic feature separated people with major depressive disorder from controls, but the authors noted the sample size and medication as limitations.
Cognitive health
Alzheimer’s disease and mild cognitive impairment
Researchers have studied pauses, fluency, articulation, prosody and higher-order spectral patterns. Some controlled datasets report high classification accuracy, but systematic reviews warn that dataset imbalance and limited external validation can inflate performance estimates.
Neurological health
Parkinson’s disease
Parkinson’s can affect phonation, articulation, pitch and rhythm. Machine-learning studies show that combinations of acoustic features can distinguish study groups, while balanced datasets and external validation remain essential for interpreting sensitivity and specificity.
Metabolic health
Type 2 diabetes
The Colive Voice study analyzed recordings from 607 U.S. adults and developed sex-specific models for predicting diabetes status. The work supports further investigation of voice as a scalable screening signal, not as a standalone diagnosis.
How to read these examples: They describe individual research studies, not the validated performance of Virtuosis AI. Results from one dataset cannot be assumed to transfer to another population, language, device or clinical setting. For details on how Virtuosis evaluates its own models, see our Clinical validation page.
What the evidence does and does not show
Voice biomarkers are a promising form of digital measurement because speech is easy to capture remotely and repeatedly. Peer-reviewed studies have reported useful associations and classification performance for specific conditions. However, performance in one study does not automatically transfer to another language, device, population or clinical setting.
Important: a voice biomarker is not automatically a diagnosis. Results may be affected by age, sex, language, microphone quality, background noise, medication, temporary illness and other health conditions. Responsible systems define the intended use, validate against an appropriate reference standard, test generalizability and communicate uncertainty.
How Virtuosis analyzes voice
1
Record
Speak naturally for at least 30 seconds through the web app, an integrated call or an API-submitted recording.
2
Measure
The system extracts acoustic patterns such as tone, pitch, pace and related signal characteristics.
3
Model
AI models compare relevant combinations of measurements for the selected use case.
4
Report
The user receives confidential health, wellbeing or communication insights appropriate to that application.
Virtuosis is designed to complement, not replace, professional judgment or established clinical assessment. Product capabilities and regulatory status depend on the specific deployment and intended use.
Privacy and responsible use
Voice recordings can be sensitive personal data. Collection should be transparent, based on an appropriate legal basis and consent process, and limited to a clearly explained purpose. Access, retention and deletion rules should be documented.
Read more about our Ethics & Privacy commitments and our Privacy Policy.
FAQ
Are “voice biomarkers” and “vocal biomarkers” the same?
Do voice biomarkers analyze what I say?
Can a voice biomarker diagnose a disease?
Is a special microphone required?
Selected scientific references
Low DM et al. Evaluation of Speech-Based Digital Biomarkers: Review and Recommendations. Digital Biomarkers. 2020.
Fagherazzi G et al. A voice-based biomarker for monitoring symptom resolution in adults with COVID-19. PLOS Digital Health. 2022.
Richard AB et al. Linguistic Markers of Subtle Cognitive Impairment in Connected Speech: A Systematic Review. JSLHR. 2024.
Briganti G, Lechien JR. Speech and Voice Quality as Digital Biomarkers in Depression: A Systematic Review. Journal of Voice. 2025.
Alqudaihi KS et al. Respiratory Diseases Diagnosis Using Audio Analysis and Artificial Intelligence: A Systematic Review. Sensors. 2024.
Hashim NW et al. Major Depressive Disorder Discrimination Using Vocal Acoustic Features. Journal of Affective Disorders. 2017.
Martínez-Nicolás I et al. Ten Years of Research on Automatic Voice and Speech Analysis of People With Alzheimer’s Disease and Mild Cognitive Impairment. Frontiers in Psychology. 2021.
Solana-Lavalle G, Rosas-Romero R. Analysis of Voice as an Assisting Tool for Detection of Parkinson’s Disease and Its Subsequent Clinical Interpretation. Biomedical Signal Processing and Control. 2021.
Elbéji A et al. A Voice-Based Algorithm Can Predict Type 2 Diabetes Status in USA Adults: Findings From the Colive Voice Study. PLOS Digital Health. 2024.
