Conference Paper

Activity Recognition on an Accelerometer Embedded Mobile Phone with Varying Positions and Orientations

DOI: 10.1007/978-3-642-16355-5_42 Conference: Ubiquitous Intelligence and Computing - 7th International Conference, UIC 2010, Xi'an, China, October 26-29, 2010. Proceedings
Source: DBLP


This paper uses accelerometer-embedded mobile phones to monitor one’s daily physical activities for sake of changing people’s
sedentary lifestyle. In contrast to the previous work of recognizing user’s physical activities by using a single accelerometer-embedded
device and placing it in a known position or fixed orientation, this paper intends to recognize the physical activities in
the natural setting where the mobile phone’s position and orientation are varying, depending on the position, material and
size of the hosting pocket. By specifying 6 pocket positions, this paper develops a SVM based classifier to recognize 7 common
physical activities. Based on 10-folder cross validation result on a 48.2 hour data set collected from 7 subjects, our solution
outperforms Yang’s solution and SHPF solution by 5~6%. By introducing an orientation insensitive sensor reading dimension,
we boost the overall F-score from 91.5% to 93.1%. With known pocket position, the overall F-score increases to 94.8%.

Full-text preview

Available from:
  • Source
    • "Toward this end, this paper proposes an online solution that fully exploits acceleration signals within the fast decision tree (DT) classifiers, without setting any predefined/fixed thresholds over any specific acceleration spaces, in order to differentiate user activities. DT-based classification is used in almost every other studies, whereas our proposed classification method provides the following novel properties, which also make findings in this paper differ from other studies such as [12]–[14], [20], [23], [27] under a similar name: 1) unsupervised learning: no priori information, no fixed thresholds, no initial training data classes; 2) adaptive: robust solution to a changing orientation of the device; 3) light-weight: efficient tree-based classification by applying sufficient signal processing techniques: no redundant computational workload; 4) online: instant context inference; 5) assisting: working standalone and/or assisting other classification algorithms by creating training data classes or input matrices; 6) updating: computationally efficient update/add/delete process on training data classes. Our proposed solution also enhances some widely used supervised classification methods, such as Gaussian mixture models (GMMs), k-NN, linear discriminant analysis (LDA), for online processing by providing training data classes without a prior offline process as well as supporting the observation analysis defined in statistical-based tools such as HMMs. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a light-weight online classification method to detect smarthpone user’s postural actions, such as sitting, standing, walking, and running. These actions are named as “user states” since they are inferred after the analysis of data acquired from the smartphones equipped accelerometer sensors. To differentiate one user state from another, many studies can be found in the literature. However, this study differs from all others by offering a computational lightweight and online classification method without knowing any priori information. Moreover, the proposed method not only provides a standalone solution in differentiation of user states, but also it assists other widely used offline supervised classification methods by automatically generating training data classes and/or input system matrices. Furthermore, we improve these existing methods for the purpose of online processing by reducing the required computational burden. Extensive experimental results show that the proposed method makes a solid differentiation in user states even when the sensor is being operated under slower sampling frequencies.
    Full-text · Article · Aug 2015
  • Source
    • "Several researchers developed AR approaches that normalize the orientation of the smartphone, resulting in an increased accuracy [8][9]. Sun et al. [10] studied AR when the location of the smartphone varied. They reported only a modest increase in the accuracy when location-specific classifiers were used, probably due to recognizing only ambulatory activities and driving, where the intensity of the phone's movement is correlated with the activity in a relatively straightforward manner. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a novel method for activity recognition and estimation of human energy expenditure with a smartphone and an optional heart-rate monitor. The method detects the presence of the devices, normalizes the orientation of the phone, detects its location on the body, and uses location-specific models to recognize the activity and estimate the energy expenditure. The normalization of the orientation and the detection of the location significantly improve the accuracy; the estimated energy expenditure is more accurate than that provided by a state-of-the-art dedicated consumer device.
    Full-text · Conference Paper · Mar 2015
  • Source
    • "In contrast, nonmulti-level approaches use shorter time windows. For example , [17] [18] [24] used a time window of 2.5 seconds, 3 seconds, and 1 second, respectively. Secondly, existing systems are based on both time domain and frequency domain features. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Although human activity recognition (HAR) has been studied extensively in the past decade, HAR on smartphones is a relatively new area. Smartphones are equipped with a variety of sensors. Fusing the data of these sensors could enable applications to recognize a large number of activities. Realizing this goal is challenging, however. Firstly, these devices are low on resources, which limits the number of sensors that can be utilized. Secondly, to achieve optimum performance efficient feature extraction, feature selection and classification methods are required. This work implements a smartphone-based HAR scheme in accordance with these requirements. Time domain features are extracted from only three smartphone sensors, and a nonlinear discriminatory approach is employed to recognize 15 activities with a high accuracy. This approach not only selects the most relevant features from each sensor for each activity but it also takes into account the differences resulting from carrying a phone at different positions. Evaluations are performed in both offline and online settings. Our comparison results show that the proposed system outperforms some previous mobile phone-based HAR systems.
    Full-text · Article · May 2014 · International Journal of Distributed Sensor Networks
Show more