I-Corps Teams: IoT Sensor Networks Detecting User Behavior in Architectural Space

这个i-Corps项目更广泛的影响/商业潜力是开发了一种经济高效的解决方案，可以在不侵犯用户隐私的情况下检测用户在建筑空间中的行为。这项技术使用被动红外(PIR)传感器来收集轻量级数据并识别用户行为，这传统上是通过摄像头和图像识别来完成的，处理能力和成本都较高。一种可能的应用包括通过生成流量和员工数量分析来优化商场市场目标和收入增长。这项技术还可以帮助商业物业和酒店更好地了解客户的需求，并通过个人行为路径预测、人群行为预测、人口预测和特定行为分析(如队列管理和购买跟踪习惯)等功能创造交叉销售机会。另一个可能的应用是通过远程协调入住和退房程序，监控清洁结果，并减少未经许可的聚会、火灾、洪水或盗窃的可能性，为房主实现租赁物业管理的自动化。这个i-Corps项目基于一种新型神经网络系统的开发，该系统可以将低像素热数据分类为用户数量和用户姿势类型。对于建筑空间中的非碎片热数据，所提出的方法能够将用户行为与无麻烦的自动化匹配。使用单个红外传感器上的热数据点，该系统可以在800ms以下的时间内将数据集实时分类为入住率和可能的姿势，准确率为95%。随着网络中传感器的增加，系统可以识别特定的用户行为和室内位置。这些行为数据是在智能生活和工作空间实现无障碍自动化的关键。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势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 I-Corps project is the development of a cost-effective solution to detect user behavior in architectural space without violating the user’s privacy. This technology uses passive infrared (PIR) sensors to collect lightweight data and identify user behavior, which is traditionally done by camera and image recognition at higher processing power and cost. One possible application includes optimizing mall market targeting and revenue growth by generating traffic flow and headcount analytics. This technology also may help commercial properties and hotels to understand their customers’ needs better and create cross-sale opportunities through features including individual behavior path predicting, crowd behavior prediction, population forecasts, and specific behavioral analytics, like queue management and purchase-tracking habits. Another possible application is in automating rental property management for homeowners by remotely coordinating check-in and check-out procedures, monitoring cleaning results, and diminishing the likelihood of unpermitted parties, fire, flood, or theft.This I-Corps project is based on the development of a novel neural network system that can classify low pixel thermal data into the number of users and type of user posture. With non-fragmented thermal data in architectural space, the proposed method is capable of matching user behavior with hassle-free automation. Using 64 thermal data points from a single PIR sensor, the system can real-time classify the dataset into occupancy numbers at a 95% accuracy rate and possible postures in under 800ms. With more sensors in the network, the system can identify specific user behavior and indoor location. These behavior data are the key to achieve frustration-free automation in smart living and working spaces.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.