By Ian DeNolfo, Executive Director, The Voice Foundation
We stand today at a profound inflection point in the long history of voice science. It is a transitional epoch where artificial intelligence (AI) has ceased to be a futuristic abstraction and has instead become an active, foundational force reshaping how researchers conduct laboratory work, how data is parsed, and how clinicians diagnose and care for patients.
As the executive director of an organization that has bridged art and medicine for more than half a century, I am constantly struck by the speed of this evolution. The human voice has served as our fundamental instrument of connection, expression, and identity for hundreds of thousands of years. It carries our emotions, exposes our vulnerabilities, and broadcasts the biological state of our bodies in ways no other physiological signal can match. Now, for the first time in history, we possess computational tools sophisticated enough to decode that immense complexity—to isolate and understand what the voice reveals about our brains, our systemic health, and our deepest well-being.
Executive Overview: A Paradigm Shift in Voice Science and Medicine
The integration of artificial intelligence into voice science is not a peripheral trend; it is an exponential transformation. While mainstream media often portrays AI in healthcare as a recent, post-pandemic phenomenon born from the explosion of large language models, the reality within our specialized field is far more historic.
For over three decades, researchers publishing in the Journal of Voice—the premier peer-reviewed publication dedicated to voice science and medicine—have quietly laid the groundwork for this revolution. What began as experimental neural networks processing basic spectral patterns in the mid-1990s has erupted into a torrent of high-impact, clinically validated machine learning applications. Today, voice is no longer viewed merely as an acoustic output of the larynx; it is increasingly recognized as a digital biomarker capable of tracking everything from neurodegenerative decline and respiratory infections to psychological distress.
This article explores the trajectory of AI in voice science: tracing its roots from early computational experiments to its current status as a cornerstone of clinical practice, examining landmark studies that have shaped the field, capturing expert perspectives on the ethical and practical implications of these tools, and outlining the road ahead for a global community of interdisciplinary researchers.
The Voice Foundation: A Legacy of Innovation and Interdisciplinary Care
This current moment of high-tech transformation does not happen in a vacuum; it builds upon a rich, half-century legacy of pioneering interdisciplinary collaboration.
In 1969, Dr. Wilbur James Gould founded The Voice Foundation in New York City. At that time, comprehensive, interdisciplinary care for the human voice was virtually nonexistent. Otolaryngologists, speech-language pathologists, vocal coaches, and performers largely operated in professional silos. Dr. Gould possessed the groundbreaking foresight to shatter those barriers, bringing together physicians, scientists, artists, and educators to share their knowledge and expertise explicitly for the care of the professional voice user.
The Foundation inaugurated its first Annual Symposium—Care of the Professional Voice—in 1972, followed by its inaugural Gala (later renamed Voices of Summer) in 1973. For more than five decades, the organization has functioned as a vital bridge between art and science, between the cold analytics of the clinical laboratory and the expressive demands of the stage.
Since 1989, The Voice Foundation has been guided by Dr. Robert Thayer Sataloff, an internationally renowned otolaryngologist who is simultaneously a professional singer and conductor. Dr. Sataloff’s staggering academic output includes more than 1,200 publications and 79 textbooks. Under his visionary leadership, the Foundation relocated its headquarters to Philadelphia and dramatically expanded its reach, deepening our scientific understanding of the voice through rigorous research and global education initiatives.
Today, our annual Symposium in Philadelphia draws hundreds of medical doctors, scientists, academic researchers, speech-language pathologists, and performing artists from every corner of the globe. Alongside this academic gathering, we continue to publish the Journal of Voice, maintaining our standing as the definitive archive of voice science and medicine.
Chronology of Computation: 30 Years of AI Research in Voice Science
To truly appreciate the current AI boom in voice research, one must look backward to realize how early our field began experimenting with machine intelligence.
The Journal of Voice has been publishing peer-reviewed research on artificial intelligence significantly longer than most people—including many tech sector workers—realize. In 1994, exactly thirty-one years ago, we published a landmark paper by Rihkanen and colleagues entitled "Spectral Pattern Recognition of Improved Voice Quality." This pioneering study utilized early neural networks to analyze voice metrics. To put that into historical context, 1994 was the exact year the World Wide Web was first emerging into public consciousness. Our academic community was publishing AI research in voice science before the vast majority of professionals had ever sent an electronic mail message.
Yet, while those early papers planted the seed, the modern era has seen exponential growth that defies conventional academic timelines. The publication data tells a striking story of acceleration:
- 1994–2015: 17 AI-related papers published over two decades.
- 2016–2019: 27 papers published as machine learning techniques matured.
- 2020–2022: 16 papers published during the global pandemic disruption.
- 2023–2025: A staggering 102 papers published in just the last three years.
To isolate the magnitude of this shift: in the year 2025 alone, the Journal of Voice published 51 AI-related papers. That single-year output surpasses our total output for the entire first twenty years of our AI publishing history combined. This is not the steady, linear growth of a traditional scientific discipline; it is an exponential curve that is fundamentally redefining what is possible in clinical diagnostics.
Supporting Context & Metrics: Beyond Citations to Real-World Impact
Academic papers are frequently judged by their citation counts, but within applied medical sciences, true impact is measured by how actively clinicians, therapists, and researchers utilize the findings in the real world. Recent usage statistics for the Journal of Voice reveal an extraordinary global appetite for AI-driven research.
On average, an AI-focused paper published in our journal is downloaded nearly 1,000 times—an extraordinarily high engagement metric for niche academic literature. However, certain landmark papers have far exceeded these baselines. For instance, a foundational study investigating machine learning applications for COVID-19 detection via voice analysis has been downloaded over 7,200 times, making it eleven times more accessed than the median article in its respective issue.
Similarly, our two most-cited AI papers—authored by Fang and colleagues (deep learning, 193 citations) and Hegde and colleagues (machine learning survey, 147 citations)—have each been downloaded nearly 5,000 times. Both were ranked among the top three most-accessed articles in their respective issues. These metrics do not represent sterile academic exercises; they represent frontline clinicians, practicing speech-language pathologists, and biomedical engineers actively searching for practical, tech-driven tools to elevate patient care.
Landmark Research Reshaping the Field
Deep Learning for Voice Pathology Detection
In 2019, Fang and collaborators published "Detection of Pathological Voice Using Cepstrum Vectors: A Deep Learning Approach." This work remains the most-cited artificial intelligence paper in our journal’s history, garnering 193 citations. The study proved conclusively that deep neural networks could identify and classify voice pathologies with remarkable precision, utilizing complex acoustic features that the human ear is entirely incapable of consciously perceiving.
This work was augmented by Hegde’s comprehensive "Survey on Machine Learning Approaches for Automatic Detection of Voice Disorders" (147 citations), which successfully mapped the early landscape of machine learning in voice science and remains mandatory reading for incoming researchers. Throughout the late 2010s and early 2020s, subsequent contributions—such as Al-Nasheri’s work on Multidimensional Voice Program parameters, Chen and Chen’s deep neural networks for voice classification, Fujimura’s one-dimensional convolutional neural networks, and Cho and Choi’s comparative analysis of CNN models for laryngoscopic images—pushed the boundaries of automated diagnostics ever further.
Voice as a Digital Biomarker
Perhaps the most transformative conceptual leap in recent years is the framing of the human voice as a digital biomarker—a non-invasive, continuous window into systemic human health.
We have published groundbreaking research demonstrating the viability of using voice analysis to detect and monitor neurodegenerative conditions such as Parkinson’s disease. A 2022 paper by Hemmerling and Wójcik-Pędziwiatr on predicting Parkinson’s severity directly from acoustic voice signals has accumulated 29 citations in just three years, paving the way for remote patient monitoring.
Beyond neurology, recent systematic reviews have explored voice quality as a digital biomarker for mental health disorders, including clinical depression and bipolar disorder. The clinical implications are profound: the prospect of screening for psychological shifts through routine vocal analysis during telehealth visits could dramatically lower barriers to entry, enable earlier therapeutic interventions, and reduce the crippling administrative burden on overstretched mental healthcare systems. Furthermore, machine learning models trained on COVID-19 vocal signatures have demonstrated how the voice can act as an early sentinel for respiratory distress.
AI Chatbots in Clinical Practice
As conversational AI has matured, our pages have mirrored the medical community’s urgent need to evaluate its safety and efficacy. In 2025, the Journal of Voice published a series of papers examining the integration of AI chatbots into clinical decision-making. Notably, a paper by Dronkers and colleagues titled "Evaluating the Potential of AI Chatbots in Treatment Decision-making for Acquired Bilateral Vocal Fold Paralysis" quickly accumulated 18 citations within its first year—an exceptional velocity indicating intense scholarly interest.
This work sparked robust academic discourse, including multiple letters to the editor and follow-up studies evaluating the precision of advanced multimodal models like ChatGPT-4o in analyzing complex laryngeal images. We are witnessing our field actively grappling with the real-time implications of artificial intelligence sitting side-by-side with human clinicians in the exam room.
Official Statements and Expert Perspectives
To understand the human element behind these computational shifts, we turn to two leading voices in our academic community who bring distinct, invaluable perspectives to the table.
Dr. Mark Berardi: AI as a Tool for Managing Complexity
Dr. Mark Berardi approaches voice science with a unique hybrid background rooted in physics and computation. He views artificial intelligence not as a replacement for human intellect, but as an essential instrument for managing biological complexity.
"Machine learning and AI are tools to help with complexity, and communication is incredibly complex in its neurobiological and physiological processes," Dr. Berardi notes. "So I think the application is entirely warranted."
His ongoing research targets voice-based digital biomarkers for aging and depression. "Acquiring speech and voice signals is relatively easy now, and we have shown potential for extracting meaningful health information," he explains. However, he cautions that the primary obstacle remains the inherent complexity of human communication itself—an obstacle that AI is uniquely suited to dismantle.
Regarding generative AI tools, Dr. Berardi offers a pragmatic, nuanced perspective. While he has not yet observed generative tools single-handedly rewriting fundamental research paradigms, he champions their ability to eliminate chronic bottlenecks in data processing and computer coding. "I can now quickly create bespoke code and edit it with natural language prompts," he says.
Intriguingly, Dr. Berardi’s lab is also studying human-AI interaction itself: how human speech patterns adapt when conversing with a chatbot versus another human, and how "synthetic" communication environments like video conference calls alter interpersonal dynamics. "We are already seeing linguistic differences in chatbot interactions," he observes, suggesting that voice science must soon expand to study the human voice not just in isolation, but in dynamic conversation with machines.
Dr. Eric Hunter: Embracing Change with Ethical Clarity
Dr. Eric Hunter focuses heavily on the institutional and structural ramifications of integrating AI into academic and clinical workflows.
"These tools aren’t just novelties," Dr. Hunter observes. "They are rapidly becoming integrated into how we conduct literature reviews, organize data, extract themes from large bodies of text, and even summarize findings."
He emphasizes an inescapable reality for modern researchers: large language models and machine learning frameworks are permanent fixtures of the scholarly ecosystem. They will inevitably become embedded within manuscript preparation, peer review, and collaborative writing infrastructure.
Rather than adopting a defensive posture, Dr. Hunter issues a vital call to action for the academic community: "Rather than resisting this shift, the critical need is to develop clear guidance around appropriate usage—for both authors and reviewers." He advocates for a transparent framework built upon four foundational pillars: accountability for generated output, absolute data integrity, protection of patient privacy, and rigorous disclosure of AI-assisted methodologies.
"Our field will benefit most," Dr. Hunter concludes, "if we embrace the productivity these tools offer while also building a shared ethical framework for their responsible use."
A Global Research Community
The rapid acceleration of AI voice research is not the isolated achievement of a few Silicon Valley tech incubators or elite medical labs; it is the product of a deeply interconnected global community.
A review of the leading authors and contributors driving scholarship within the Journal of Voice reveals an international coalition of brilliant minds. Scholars such as Jérôme René Lechien, Dimitar Deliyski, Stephanie Zacharias, and Ahmed Yousef—alongside Paavo Alku, Leonardo Wanderley Lopes, and Maryam Naghibolhosseini—have spearheaded dozens of critical studies. Alongside hundreds of co-investigators based at universities and medical centers spanning every inhabited continent, these researchers are defining the modern era of computational voice science.
This global collaboration proves that our community is united not by geography, but by a shared commitment to unlocking the mysteries of human vocal production through advanced computational methods.
The Path Forward: Challenges and Action Items
As we look toward the next decade, the Journal of Voice will remain the premier global platform for this rapidly evolving discipline. We remain steadfastly committed to publishing rigorous, peer-reviewed research that pushes the boundaries of human knowledge while maintaining the unyielding standards of scientific integrity upon which Dr. Gould founded our institution.
However, navigating this brave new world requires collective action across three primary fronts:
- Thoughtful Engagement: We must engage with AI tools consciously. These systems are designed to manage complexity and augment our capabilities; they are instruments to amplify human expertise, never to replace clinical judgment.
- Ethical Harmonization: We must actively participate in establishing shared ethical guidelines. Dr. Hunter’s call for a cohesive ethical framework cannot be answered by a single journal or university; it demands the active cooperation of our entire global field.
- Open Collaboration: We must continue to publish, debate, and share our findings. The 161 AI-focused papers published in our journal to date represent merely the foundation. Every clinician, speech-language pathologist, researcher, and educator holds unique insights that can drive our collective understanding forward.
An Extraordinary Moment in Time
For hundreds of thousands of years, the human voice has carried our triumphs, our griefs, our art, and our health across time and space. It is the most intimate and revealing physiological signal we possess.
Now, for the first time in human history, we have engineered tools sophisticated enough to truly comprehend that complexity—tools capable of decoding what the voice whispers about our neurological health, detecting subtle pathologies before they manifest as crisis, and democratizing access to specialized voice care for underserved populations worldwide.
This is an extraordinary moment in the history of medicine and art. The Journal of Voice and The Foundation stand proudly at the absolute center of this transformation. The coming decade promises even greater discoveries, and we look forward to charting that uncharted territory alongside our global community.
About the Author
Ian DeNolfo is the Executive Director of The Voice Foundation, which publishes the peer-reviewed Journal of Voice. A graduate of The Juilliard School and The Curtis Institute of Music, he previously performed as a leading tenor at major opera houses worldwide before transitioning his career toward arts administration, scientific advocacy, and institutional leadership within The Voice Foundation.
