The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to transform voice intonations into voice biomarkers to predict the existence of disease, monitor progression or deterioration of disease or chronic condition, predict hospitalization and mortality, focus on patients in need, and as a result, optimize care and cost. Stress, anxiety, and depression cost American employers an estimated $500 B annually in lost productivity. Furthermore, eight risks and behaviors associated with mental health drive 15 chronic conditions, accounting for 80% the total costs for all chronic illnesses worldwide and representing a projected $47 T problem by 2030. Voice signals indicate a variety of health conditions, emotions, and diseases. While wearables are becoming a ubiquitous tool to assess physical variables, their ability to measure psychological variables remains limited.The company has developed a neural-network model specifically to analyze raw text and audio from natural conversation, discovering speech patterns indicative of depression. The company is advancing research on human emotion classifiers, the first of its study across international geographies; this project will combine sensor inputs for state-of-the-art machine learning language-based models, design a hyper-individualized behavioral recommendation system for stress triggers, and develop a quantitative measurement on mental health that is both scalable and personalized to address the marketplace gap between simple apps and advanced neuropsychiatric treatment.
This Small Business Innovation Research (SBIR) Phase I project is dedicated to building a smart voice journaling platform utilizing voice biomarkers to measure and predict well-being. The major research objectives in this proposal include (1) intuitive human-computer voice interactions through various smart devices including phones, earbuds, watches, home, and in-car audio, (2) developing, training, refining, and scaling custom neural networks, (3) creating deep reinforcement learning models to serve relevant recommendations and actions to users, and (4) building visual representations of progress from individual journal entries. The anticipated outcome of this innovation is a personalized deep learning-based system that can be scaled to smart devices on-demand and robust enough to cover diverse, multicultural backgrounds worldwide. This company is using sensors to deliver objective measurements, algorithms to support the physician and psychologist in their assessments and care delivery, and non-pharmacological health-supportive tools as an emerging category of digital therapeutics.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
