IPSN : [proceedings]. IPSN (Conference)最新文献

The increased availability of on-body sensors gives researchers access to rich time-series data, many of which are related to human health conditions. Sharing such data can allow cross-institutional collaborations that create advanced data-driven models to make inferences on human well-being. However, such data are usually considered privacy-sensitive, and publicly sharing this data may incur significant privacy concerns. In this work, we seek to protect clinical time-series data against membership inference attacks, while maximally retaining the data utility. We achieve this by adding an imperceptible noise to the raw data. Known as adversarial perturbations, the noise is specially trained to force a deep learning model to make inference mistakes (in our case, mispredicting user identities). Our preliminary results show that our solution can better protect the data from membership inference attacks than the baselines, while succeeding in all the designed data quality checks.

We present a deep learning framework for fast 3D activity localization and tracking in a dynamic and crowded real world setting. Our training approach reverses the traditional activity localization approach, which first estimates the possible location of activities and then predicts their occurrence. Instead, we first trained a deep convolutional neural network for activity recognition using depth video and RFID data as input, and then used the activation maps of the network to locate the recognized activity in the 3D space. Our system achieved around 20cm average localization error (in a 4m × 5m room) which is comparable to Kinect's body skeleton tracking error (10-20cm), but our system tracks activities instead of Kinect's location of people.

A variety of health and behavioral states can potentially be inferred from physiological measurements that can now be collected in the natural free-living environment. The major challenge, however, is to develop computational models for automated detection of health events that can work reliably in the natural field environment. In this paper, we develop a physiologically-informed model to automatically detect drug (cocaine) use events in the free-living environment of participants from their electrocardiogram (ECG) measurements. The key to reliably detecting drug use events in the field is to incorporate the knowledge of autonomic nervous system (ANS) behavior in the model development so as to decompose the activation effect of cocaine from the natural recovery behavior of the parasympathetic nervous system (after an episode of physical activity). We collect 89 days of data from 9 active drug users in two residential lab environments and 922 days of data from 42 active drug users in the field environment, for a total of 11,283 hours. We develop a model that tracks the natural recovery by the parasympathetic nervous system and then estimates the dampening caused to the recovery by the activation of the sympathetic nervous system due to cocaine. We develop efficient methods to screen and clean the ECG time series data and extract candidate windows to assess for potential drug use. We then apply our model on the recovery segments from these windows. Our model achieves 100% true positive rate while keeping the false positive rate to 0.87/day over (9+ hours/day of) lab data and to 1.13/day over (11+ hours/day of) field data.