Citation Context

...ion can be measured by phones. Thus, human-centric environmental sensing is brought forward. Based on smartphones, we are able to know and record user movements (standing, sitting, walking, running) [=-=Park et al. 2012-=-; Brajdic and Harle 2013; Wu et al. 2013b], user environments (indoor/outdoor, office/home, on board) [Rachuri et al. 2010; Zhou et al. 2012b; Yang et al. 2014], and user activities (driving, cycling,...

Citation Context

...nition has been extended from distinguishing not only between the user moving and standing still, but to further include estimating the walking speed, climbing on stairs, taking an elevator, and more =-=[24, 27]-=-. Gusenbauer et al. [9], for instance, use detected activities to constrain the position estimate to known locations of stairs and elevators. While the near-ubiquity of WiFi access points enables abso...

Citation Context

...which corresponds to a window size of only 1/6th of the ‘natural’ window. Similarly, windows 1 covering more than than 24 hours are imaginable, such as those related to the nutrition over the last 48 hours. Clearly, a fixed window size based on the mentioned proportion will not guarantee optimal results, and we will have to include and optimize the size of the window as a parameter in the definition of features. Although a feature construction step (even including aggregation) is the backbone of many activity recognition projects [5], all too often this step is presented as a one-pass process [3, 16, 18, 19], such that only a fixed set of features becomes available for the actual modeling step. The resulting features are static and constructed manually, either based on some domain knowledge about the physics involved, or by making default choices. It is not hard to imagine that this step is, in fact, the result of several iterations of trial and error. Moreover, this fixed set is required to be relatively small, for reasons of memory or storage. The iterative method we propose, Accordion, is an embedded approach that does both feature construction (building candidate aggregate features), and feat...

Citation Context

...ollowing equation to estimate the metabolic consumption of walking [9]: V O2 = 3.5 + 0.1(speed) + 1.8(speed)(incline). (1) The speed parameter of this equation can be estimated from the accelerometer =-=[7]-=-. Finally, the metabolic estimation (V O2) is mapped to caloric consumption via the following equation [2]: E = (V O2 −BMR) ∗ 5/(Body Mass). (2) BMR in Equation (2) is the basal metabolic rate. Conclu...

Citation Context

...est their method (first walking detected) on my dataset. Sensors 2015, 15 25502 Table 16. Comparison with other accelerometer-based phone placement solutions. LOSO, leave-one-subject-out. Park et al. =-=[19]-=- Wiese et al. [8] Fujinami et al. [20] Alanezi et al. [22] Wahl et al. [25] Kunze et al. [9] Mannini et al. [18] This Work Positions 4: Bag, ear, hand, pocket 4: Pocket, bag, hand, out 9: Neck, chest,...

Citation Context

...ees, denoted as (x, y), must satisfy (x − x0)2 + (y − y0)2 ≤ (V ∆T + D)2. Clearly this is the most simple and conservative estimation, which can be significantly improved by considering walking speed =-=[9]-=-, direction [19], and other coarse-grained localization techniques by leveraging various sensors. The bootstrapping phase will be triggered manually or automatically, for example, after sensing the us...

Citation Context

...ding the above SVM classifications (technically, their distance from the hyperplane) through an HMM [5]. Further improvements could come from the use of more sophisticated audio [6] and accelerometer =-=[26]-=- features. 55 56 Chapter 4 Sightless Image Labeling We seek to provide a (possibly blind) wearer with the ability to teach their wearable system about the acquaintances they’d like it to subsequently ...

Citation Context

...point start/stop times ofspuffing/smoking. Wrist or arm-worn accelerometers have been used successfully for detection of homogeneoussand heterogeneous activities, such as walking, drinking and eating =-=[8, 19]-=-. Wrist-worn accelerometers may beseasier to wear, more informative, more energy efficient and more affordable than other options for some detectionstasks.sTable 1 summarizes prior work on puffing/smo...

Citation Context

...ding the above SVM classifications (technically, their distance from the hyperplane) through an HMM [5]. Further improvements could come from the use of more sophisticated audio [6] and accelerometer =-=[26]-=- features. 55 56 Chapter 4 Sightless Image Labeling We seek to provide a (possibly blind) wearer with the ability to teach their wearable system about the acquaintances they'd like it to subsequently ...