Assessment of food intake using speech-understanding technology

描述(由申请者提供)：肥胖症在美国和世界范围内的流行率保持在流行水平。目前，由于缺乏准确、低负担的食物摄入量自我评估方法，肥胖研究和预防和治疗肥胖症的有效公共卫生方法的进展受到很大阻碍。这个翻译研究项目将开发一个原型多模式计划，旨在成为一种自我评估能量和营养摄入量的最先进方法。该计划将允许用户与定制设计的口语对话系统进行互动，以获取足够的信息来确定特定食物的定量数量，根据这些信息，将使用已公布的食物成分数据来计算能量和营养摄入量的详细报告。我们将使用之前开发的WAMI(Web可访问多模式接口)框架作为底层基础设施。该原型将在成年男性和女性身上进行测试，他们将对自己的食物摄入量进行录音，并在7天内同时使用新程序。新方法和基于录音的传统计算方法将比较新方法和传统计算方法所消耗的食物类型和数量、评估每日摄入量所需的时间以及计算出的能量摄入量。预计原型将证明进一步的开发工作和创建一个比现有方法更准确、用户负担更低的程序是合理的。新计划新颖且极具创新性，因为它将包括：1)首次使用语音理解技术作为多模式对话互动的一部分，以减少用户在准确识别所消费食物时的负担；2)首次使用自然口语理解技术，以进一步减少用户负担；以及3)增加选择，以提高准确性和减少用户负担，包括用户对食物成分的个性化。该项目是非常可行的，因为多学科团队在开发多模式和语音理解软件以及验证食物摄取方法方面拥有广泛的专业知识，并已获得与应用相关的初步数据。

DESCRIPTION (provided by applicant): The prevalence of obesity remains at epidemic levels in the U.S. and worldwide. Currently, research on obesity and advances in effective public health approaches to prevention and treatment of obesity are substantially hampered by the lack of accurate, low burden methods for self-assessment of food intake. This translational research project will develop a prototype multimodal program designed to become a state-of-the- art method for self-assessment of energy and nutrient intakes. The program will allow users to interact with a custom-designed spoken dialogue system to solicit sufficient information to determine quantitative amounts of specific food items, from which detailed reports of energy and nutrient intakes will be calculated using published data on food composition. We will use our previously developed WAMI (Web-accessible Multimodal Interface) framework as the underlying infrastructure. The prototype will be tested in adult men and women, who will make audio recordings of their food intake and simultaneously use the new program over 7 days. Types and amounts of foods consumed, time taken to assess daily intake, and calculated energy intake will be compared between the new method and conventional calculations based on the audio recording. The prototype is expected to justify further developmental work and creation of a program that is both more accurate and with lower user burden that existing methods. The new program is novel and highly innovative because it will include: 1) the first use of speech understanding technology as part of multimodal dialogue interaction to reduce user burden during accurate identification of consumed foods; 2) first use of natural spoken language understanding technology to further reduce user burden; and 3) additional options to enhance accuracy and reduce user burden including individualization of food composition by users. The project is very feasible because the multidisciplinary team has extensive expertise in both the development of multimodal and speech understanding software and validation of food intake methods, and has also obtained preliminary data relevant to the application.