Feature extraction from smartphone inertial signals for human activity segmentation

San-Segundo, Rubén; Montero, Juan Manuel; Barra-Chicote, Roberto; Fernández, Fernando; Pardo, José Manuel

doi:10.1016/j.sigpro.2015.09.029

Cited by 88 publications

(60 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In [27], the authors approach the problem of acceleration-based activity recognition using strategies that are typical in speech processing, such as Mel Frequency Cepstral Coefficientss (MFCCs) and Perceptual Linear Predictions (PLPs) coefficients. Their framework extracts a total of 561 time-domain and frequency-domain features.…”

Section: Survey Of Feature Extraction Pipelinesmentioning

confidence: 99%

A Comprehensive Study of Activity Recognition Using Accelerometers

et al. 2018

View full text Add to dashboard Cite

This paper serves as a survey and empirical evaluation of the state-of-the-art in activity recognition methods using accelerometers. The paper is particularly focused on long-term activity recognition in real-world settings. In these environments, data collection is not a trivial matter; thus, there are performance trade-offs between prediction accuracy, which is not the sole system objective, and keeping the maintenance overhead at minimum levels. We examine research that has focused on the selection of activities, the features that are extracted from the accelerometer data, the segmentation of the time-series data, the locations of accelerometers, the selection and configuration trade-offs, the test/retest reliability, and the generalisation performance. Furthermore, we study these questions from an experimental platform and show, somewhat surprisingly, that many disparate experimental configurations yield comparable predictive performance on testing data. Our understanding of these results is that the experimental setup directly and indirectly defines a pathway for context to be delivered to the classifier, and that, in some settings, certain configurations are more optimal than alternatives. We conclude by identifying how the main results of this work can be used in practice, specifically in experimental configurations in challenging experimental conditions.

show abstract

Section: Survey Of Feature Extraction Pipelinesmentioning

confidence: 99%

A Comprehensive Study of Activity Recognition Using Accelerometers

et al. 2018

View full text Add to dashboard Cite

show abstract

“…Falling however, is an activity with a unique motion that is hard to mix up with other activities of everyday living, thus these methods may not fully work in our scenario. Finally, for classification, classifiers that are commonly used are Decision Trees [18], Hidden Markov Models [21], and kNN [22] and recently Neural Networks [23]. In our work, we rely on a sliding window approach, similar to [2].…”

Section: Activity Recognition Based On Inertial Datamentioning

confidence: 99%

Vision and Acceleration Modalities: Partners for Recognizing Complex Activities

Diete

Sztyler

Stuckenschmidt

2019

2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops)

View full text Add to dashboard Cite

Wearable devices have been used widely for human activity recognition in the field of pervasive computing. One big area of in this research is the recognition of activities of daily living where especially inertial and interaction sensors like RFID tags and scanners have been used. An issue that may arise when using interaction sensors is a lack of certainty. A positive signal from an interaction sensor is not necessarily caused by a performed activity e.g. when an object is only touched but no interaction occurred afterwards. In our work, we aim to overcome this limitation and present a multi-modal egocentricbased activity recognition approach which is able to recognize the critical activities by looking at movement and object information at the same time. We present our results of combining inertial and video features to recognize human activities on different types of scenarios where we achieve a F1-measure up to 79.6%.

show abstract

“…We utilized Extremely Randomized Trees (ET) [22], and Random Forests (RF) [23] as examples of tree-based approaches, Multilayer Perceptrons (MLP), and Support Vector Machines (SVM) [24] as examples of statistical and connectionist approaches. The choice of these classifiers is based on previous research that showed that they consistently performed successfully in activity recognition tasks with different datasets [1], [3], [10]. The complexity of any classification problem is inversely correlated with the success of feature selection and extraction.…”

Section: Classificationmentioning

confidence: 99%

“…Activity recognition from smart-phone sensors can be further divided into systems that deal with the problem as a temporal time-series classification task employing approaches like HMMs [1], and systems that treat activity recognition as a standard classification problem from predefined windows of sensory data [2]- [4]. This paper focuses on feature extraction and selection common to both approaches.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, the proposed method further reduces expected energy requirements of the algorithm by not requiring continuous sampling of accelerometer data. This means that the algorithms used for classification can only assume the availability of short windows of data that are not related temporally which eliminates algorithms that try to model the temporal structure like HMMs and LSTMs [1], [5], [6] The main goal of this work is to devise an activity recognition system that achieves consistent state-of-the-art results on a wide range of benchmark datasets using a single sensor (accelerometer) without the need to continuously sample that sensor (sparse sampling).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Selecting Orientation-Insensitive Features for Activity Recognition from Accelerometers

Mohammad

Matsumoto

Hoashi

2019

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

Activity recognition from sensors is a classification problem over time-series data. Some research in the area utilize time and frequency domain handcrafted features that differ between datasets. Another categorically different approach is to use deep learning methods for feature learning. This paper explores a middle ground in which an off-theshelf feature extractor is used to generate a large number of candidate timedomain features followed by a feature selector that was designed to reduce the bias toward specific classification techniques. Moreover, this paper advocates the use of features that are mostly insensitive to sensor orientation and show their applicability to the activity recognition problem. The proposed approach is evaluated using six different publicly available datasets collected under various conditions using different experimental protocols and shows comparable or higher accuracy than state-of-the-art methods on most datasets but usually using an order of magnitude fewer features.

show abstract

Feature extraction from smartphone inertial signals for human activity segmentation

Cited by 88 publications

References 27 publications

A Comprehensive Study of Activity Recognition Using Accelerometers

A Comprehensive Study of Activity Recognition Using Accelerometers

Vision and Acceleration Modalities: Partners for Recognizing Complex Activities

Selecting Orientation-Insensitive Features for Activity Recognition from Accelerometers

Contact Info

Product

Resources

About