Fluid intake is an important information for many health and assisted living applications. At the same time it is inherently difficult to monitor. Existing reliable solutions require augmented drinking containers, which severely limits the applicability of such systems. In this paper we investigate two key components of an unobtrusive, wearable solution that is independent of a particular drinking container or environment.We first describe a system for spotting individual instances of drinking (lifting a container to the mouth and taking a single sip) in a continuous stream of data from a wrist-worn acceleration sensor. We show that drinking motion can be detected across different drinking containers (glass, cup, large beer mug, bottle) on a large dataset (560 drinking motion instances from six users, embedded in 5.84 hours of complex natural activities). An average performance of 84% recall at 94% precision was achieved for the drinking motion spotting.Based on the events derived from drinking event spotting, we show how additional information can be obtained. Specifically, we demonstrate the recognition of container types and fluid level from upper body postures during drinking events. Nine containers and three container fluid levels were evaluated to recognize container type and fluid amounts with three users.Recognition rate for container type was 75%, and for fluid level 72%.
Abstract-Choosing the right feature for motion based activity spotting is not a trivial task. Often, features derived by intuition or that proved to work well in previous work are used. While feature selection algorithms allow automatic decision, definition of features remains a manual task. We conduct a comparative study of features with very different origin. To this end, we propose a new type of features based on polynomial approximation of signals. The new feature type is compared to features used routinely for motion based activity recognition as well as to recently proposed body-model based features. Experiments were performed on three different, large datasets allowing a thorough, in-depth analysis. They not only show the respective strengths of the different feature types but also their complementarity resulting in improved performance through combination. It shows that each feature type with its individual and complementary strengths and weaknesses can improve results by combination.
We describe the use of upper leg mounted force sensitive resistors (FSR) to analyze muscle activity during bicycling. We demonstrate that FSRs can provide information that is not accessible to motion sensors, like the gear in which a person is cycling or rather the amount of force applied to the pedals. This is exemplary for many other activities where the effort and subtle muscle activities patterns play a role in determining how well an activity is being performed.Together with a sports clothing manufacturer (FALKE AG) we have developed special shorts for the integration of FSRs. We have developed and implemented an online recognition method. This method has been evaluated in an elaborate experiment.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.