SBIR Phase I: Scaling Mental Healthcare in COVID-19 with Voice Biomarkers

这个小企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是使用语音作为心理健康的实时测量。将语音语调转化为生物标志物可以促进疾病诊断和进展，从而支持价值130亿美元的虚拟医疗保健行业，该行业在COVID-19之前每年增长27%。此外，同伴对精神卫生的支持增加了自我保健的参与，减少了药物使用和抑郁，特别是对弱势群体而言。该项目将推动机器学习在群体环境中语音心理健康生物标志物的使用。这个小企业创新研究（SBIR）第一阶段项目将为基于深度强化学习的系统定义语音生物标志物特征。该项目将推进一种语音生物标记技术，可以作为快速的行为健康诊断，有可能取代目前基于纸张的PHQ-9和GAD-7测试。优先考虑的是基于奖励功能的群体治疗中个体和活动的最佳组合，以最大限度地改善抑郁和焦虑得分。主要的技术挑战包括：(1)捕捉视频中的非语言线索；(2)解译多扬声器音频处理；(3)建立深度强化学习模型，服务于相关的分组匹配和后续练习；(4)从小组会议中建立引人入胜的视觉反馈。这项创新的预期技术成果将是在临床相关环境中为基于深度强化学习的系统定义语音生物标志物特征和奖励功能，以改善抑郁和焦虑的治疗效果。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to use voice as a real-time measurement of mental health. Transforming voice intonations into biomarkers could enable disease diagnosis and progression, supporting the $13 B virtual health care sector that was growing 27% annually prior to COVID-19. Furthermore, peer support for mental health increases engagement in self-care decreases substance use and depression, particularly for vulnerable populations. The project will advance the use of machine learning for voice mental health biomarkers in a group setting. This Small Business Innovation Research (SBIR) Phase I project will define voice biomarker features for a deep reinforcement learning based system. This project will advance a voice biomarker technology that can serve as fast behavioral health diagnostic, potentially superseding the current paper-based PHQ-9 and GAD-7 tests. The priority is to scale the optimal mix of individuals and activities for group therapy based on reward functions that maximize improvements in depression and anxiety scores. The major technical challenges include: (1) capturing nonverbal cues in a video; (2) interpreting multi-speaker audio processing; (3) creating deep reinforcement learning models to serve relevant group matches and follow-up exercises; and (4) building engaging visual feedback of progress from group meetings. The anticipated technical result of this innovation will be to define voice biomarker features and reward functions for a deep reinforcement learning based system in clinically relevant settings to improve depression and anxiety treatment outcomes.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.