Context Recognition In-the-Wild: Unified Model for Multi-Modal Sensors and Multi-Label Classification
The DOI matches its Crossref record; 24 of the 24 references it cites were found in a registry or an index.
Abstract
Automatic recognition of behavioral context (location, activities, body-posture etc.) can serve health monitoring, aging care, and many other domains. Recognizing context in-the-wild is challenging because of great variability in behavioral patterns, and it requires a complex mapping from sensor features to predicted labels. Data collected in-the-wild may be unbalanced and incomplete, with cases of missing labels or missing sensors. We propose using the multiple layer perceptron (MLP) as a multi-task model for context recognition. Based on features from multi-modal sensors, the model simultaneously predicts many diverse context labels. We analyze the advantages of the model's hidden layers, which are shared among all sensors and all labels, and provide insight to the behavioral patterns that these hidden layers may capture. We demonstrate how recognition of new labels can be improved when utilizing a model that was trained for an initial set of labels, and show how to train the model to withstand missing sensors. We evaluate context recognition on the previously published ExtraSensory Dataset, which was collected in-the-wild. Compared to previously suggested models, the MLP improves recognition, even with fewer parameters than a linear model. The ability to train a good model using data that has incomplete, unbalanced labeling and missing sensors encourages further research with uncontrolled, in-the-wild behavior.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
(2014) · Cited by 33,419
Activity recognition using cell phone accelerometers
Kwapisz, J. R., Weiss, G. M., & Moore, S. A. (2011) · ACM SIGKDD Explorations Newsletter · doi:10.1145/1964897.1964918 · Cited by 2,912 · cited 3×
“[7] targeted six body states and used a built-in accelerometer in a smartphone that was placed in the participants’ front pant pocket.”
Method(2011) · Cited by 2,752
The Balanced Accuracy and Its Posterior Distribution
Brodersen, K. H., Ong, C. S., Stephan, K. E., & Buhmann, J. M. (2010) · 2010 20th International Conference on Pattern Recognition · doi:10.1109/icpr.2010.764 · Cited by 1,816 · cited 1×
“A convenient way to consider both sensitivity and specificity is simply to average them, resulting in the balanced accuracy metric [1], whose chance level is always 0.”
BackgroundProceedings of the 29th International Conference on Machine Learning (ICML-12)
Langford, J. C., & Pineau, J. (2012) · arXiv (Cornell University) · doi:10.48550/arxiv.1207.4676 · Cited by 1,618
Ermes, M., Parkka, J., Mantyjarvi, J., & Korhonen, I. (2008) · IEEE Transactions on Information Technology in Biomedicine · doi:10.1109/titb.2007.899496 · Cited by 707 · cited 3×
“[3] addressed some aspects of in-the-wild behavior.”
BackgroundMethodDetecting activities of daily living in first-person camera views
Pirsiavash, H., & Ramanan, D. (2012) · 2012 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2012.6248010 · Cited by 649 · cited 7×
“We perform the action-recognition task that they describe in [15], where each item is a pre-segmented clip annotated with a single action out of eighteen.”
BackgroundMethodInfluentialOptimal Thresholding of Classifiers to Maximize F1 Measure
Lipton, Z. C., Elkan, C., & Naryanaswamy, B. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-662-44851-9_15 · Cited by 596 · cited 1×
“This effect can result in misleading conclusions when comparing two systems [9].”
ResultRecognizing Detailed Human Context in the Wild from Smartphones and Smartwatches
Vaizman, Y., Ellis, K., & Lanckriet, G. (2017) · IEEE Pervasive Computing · doi:10.1109/mprv.2017.3971131 · Cited by 324
Self-citationRecognizing human motion with multiple acceleration sensors
Mantyjarvi, J., Himberg, J., & Seppanen, T. (2002) · 2001 IEEE International Conference on Systems, Man and Cybernetics. e-Systems and e-Man for Cybernetics in Cyberspace (Cat.No.01CH37236) · doi:10.1109/icsmc.2001.973004 · Cited by 323 · cited 3×
“Previous studies used MLP for the multi-class setting, where the purpose was selecting a single activity (the one with highest probability output) out of a set of mutually-exclusive options [3, 4, 7, 12, 16].”
MethodFeature Selection and Activity Recognition from Wearable Sensors
Pirttikangas, S., Fujinami, K., & Nakajima, T. (2006) · Lecture Notes in Computer Science · doi:10.1007/11890348_39 · Cited by 229 · cited 5×
“Unlike kNN [16, 19], MLP has a model size (and test run-time) that”
BackgroundMethodInfluentialPerformance metrics for activity recognition
Ward, J. A., Lukowicz, P., & Gellersen, H. W. (2011) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/1889681.1889687 · Cited by 182 · cited 1×
“[22] addressed these issues and suggested metrics that partition false negative rate (1-sensitivity) and false positive rate (1-specificity) to describe finer types of errors.”
BackgroundModeling Missing Data in Clinical Time Series with RNNs
Lipton, Z. C., Kale, D. C., & Wetzel, R. C. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1606.04130 · Cited by 159
Activity Recognition on Smartphones via Sensor-Fusion and KDA-Based SVMs
Khan, A. M., Tufail, A., Khattak, A. M., & Laine, T. H. (2014) · International Journal of Distributed Sensor Networks · doi:10.1155/2014/503291 · Cited by 150 · cited 3×
“This is different from the two-stage approach in [6], where they first used KDA to learn a representation, and then trained an SVM classifier.”
BackgroundMethodInfluentialTowards detection of bad habits by fusing smartphone and smartwatch sensors
Shoaib, M., Bosch, S., Scholten, H., Havinga, P. J. M., & Incel, O. D. (2015) · 2015 IEEE International Conference on Pervasive Computing and Communication Workshops (PerCom Workshops) · doi:10.1109/percomw.2015.7134104 · Cited by 114 · cited 4×
“[19] started with “simple” activities (body movements) and then extended their experiments to “complex” activities (e.”
BackgroundMethodInfluentialSensor-based observations of daily living for aging in place
Lee, M. L., & Dey, A. K. (2014) · Personal and Ubiquitous Computing · doi:10.1007/s00779-014-0810-3 · Cited by 111 · cited 1×
“Automatic recognition of behavioral context can serve many applications, such as monitoring physical activity [2], logging older adults’ functional independence to promote aging at home [8], and context-adaptive personal assistant systems.”
BackgroundMulti-Sensor Fusion for Enhanced Contextual Awareness of Everyday Activities with Ubiquitous Devices
Guiry, J., Van de Ven, P., & Nelson, J. (2014) · Sensors · doi:10.3390/s140305687 · Cited by 91
Human Activity Detection in Smart Home Environment with Self-Adaptive Neural Networks
Zheng, H., Wang, H., & Black, N. D. (2008) · IEEE International Conference on Networking, Sensing and Control (ICNSC) · doi:10.1109/icnsc.2008.4525459 · Cited by 90 · cited 4×
“Unlike the unsupervised discovery of new contexts in [18, 23], we stay in the realm of supervised learning.”
BackgroundMethodInfluentialObjective Assessment of Physical Activity: Classifiers for Public Health
KERR, J., PATTERSON, R. E., ELLIS, K., GODBOLE, S., JOHNSON, E., LANCKRIET, G., & STAUDENMAYER, J. (2016) · Medicine & Science in Sports & Exercise · doi:10.1249/mss.0000000000000841 · Cited by 79
Self-citationAmbientSense: A real-time ambient sound recognition system for smartphones
Rossi, M., Feese, S., Amft, O., Braune, N., Martis, S., & Tröster, G. (2013) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomw.2013.6529487 · Cited by 62 · cited 1×
“[17] collected sound clips, available on the web, that were annotated with tags describing locations (e.”
BackgroundNatarajan, A., Angarita, G. A., Gaiser, E., Malison, R. T., Ganesan, D., & Marlin, B. M. (2016) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/2971648.2971666 · Cited by 48 · cited 2×
“[13] worked on using wearable Electrocardiography sensors to detect cocaine use.”
MethodMulti-sensor physical activity recognition in free-living
Ellis, K., Kerr, J., Godbole, S., & Lanckriet, G. (2014) · Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing: Adjunct Publication · doi:10.1145/2638728.2641673 · Cited by 44 · cited 1×
“Automatic recognition of behavioral context can serve many applications, such as monitoring physical activity [2], logging older adults’ functional independence to promote aging at home [8], and context-adaptive personal assistant systems.”
BackgroundSelf-citationDiscovery of activity composites using topic models: An analysis of unsupervised methods
Seiter, J., Amft, O., Rossi, M., & Tröster, G. (2014) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2014.05.007 · Cited by 37 · cited 4×
“[18] described using topic modeling to discover new contexts — common temporal-sequences of “activity primitives”.”
BackgroundInfluentialNovel neural network based fusion for multistream ASR
Mallidi, S. H., & Heřmanský, H. (2016) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2016.7472765 · Cited by 29 · cited 4×
“Our treatment of missing sensors is similar to the stream-dropout used in [11] or the simple zero-imputation used in [10], but we make sure to normalize the total contribution of the available sensors at every example.”
BackgroundMethodInfluential
Research works printed but not linked: 1
- 10.5555/2627435.2670313 · doi:10.5555/2627435.2670313
Cited by
Robust human locomotion and localization activity recognition over multisensory
Khan, D., Alonazi, M., Abdelhaq, M., Mudawi, N. A., Algarni, A. M., Jalal, A., & Liu, H. (2024) · Frontiers in Physiology · doi:10.3389/fphys.2024.1344887 · Cited by 220
Cruciani, F., Vafeiadis, A., Nugent, C., Cleland, I., McCullagh, P., Votis, K., Giakoumis, D., Tzovaras, D., Chen, L., & Hamzaoui, R. (2020) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-020-00026-2 · Cited by 142
“The data-set, however, contained recordings where the feature vector per recording was larger than the (≈ 430 ) × 13 , as described in the original paper (Vaizman 2017).”
MethodInternet of Things (IoT) Based Activity Recognition Strategies in Smart Homes: A Review
Babangida, L., Perumal, T., Mustapha, N., & Yaakob, R. (2022) · IEEE Sensors Journal · doi:10.1109/jsen.2022.3161797 · Cited by 149
ExtraSensory App: Data Collection In-the-Wild with Rich User Interface to Self-Report Behavior
Vaizman, Y., Ellis, K., Lanckriet, G., & Weibel, N. (2018) · Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3173574.3174128 · Cited by 115
Self-citationColloSSL: Collaborative Self-Supervised Learning for Human Activity Recognition
Jain, Y., Tang, C. I., Min, C., Kawsar, F., & Mathur, A. (2022) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3517246 · Cited by 113
Metadata
- Authors
- Vaizman, Y., Weibel, N., & Lanckriet, G.
- Affiliations
- University of California San Diego (US)
- Venue
- Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Published
- 2018-01-08
- Volume
- 1
- Issue
- 4
- Pages
- 1-22
- ISSN
- 2474-9567 (ISSN-L)
- OpenAlex
- W2783920628
- Cited by
- 122
- References
- 24
- Citations by year
- 2026: 7 · 2025: 18 · 2024: 15 · 2023: 13 · 2022: 15 · 2021: 13 · 2020: 19 · 2019: 16 · 2018: 6
- FWCI
- 6.76
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Social Sciences › Transportation › Human Mobility and Location-Based Analysis
- Physical Sciences › Computer Science › Computer Science Applications › Mobile Crowdsensing and Crowdsourcing
- Keywords
- context recognition, in-the-wild, multi-task learning, multilayer perceptron, multimodal sensor fusion, missing labels, class imbalance, health monitoring, elderly care, activity recognition, posture recognition
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.