Abstract
This paper addresses one of the main challenges in physical activity monitoring, as indicated by recent benchmark results: The difficulty of the complex classification problems exceeds the potential of existing classifiers. Therefore, this paper proposes the ConfAdaBoost.M1 algorithm. This algorithm is a variant of the AdaBoost.M1 that incorporates well-established ideas for confidence-based boosting. ConfAdaBoost.M1 is compared to the most commonly used boosting methods using benchmark datasets from the UCI machine learning repository. Moreover, it is evaluated on an activity recognition and an intensity estimation problem, including a large number of physical activities from the recently released PAMAP2 dataset. The presented results indicate that the proposed ConfAdaBoost.M1 algorithm significantly improves the classification performance on most of the evaluated datasets, especially for larger and more complex classification tasks. Finally, two empirical studies are designed and carried out to investigate the feasibility of ConfAdaBoost.M1 for physical activity monitoring applications in mobile systems.
Similar content being viewed by others
Notes
The remaining 3 activities from the dataset are discarded for the following reasons: drive car contains data from only one subject, while watch TV and computer work are not considered due to their high resemblance to the sit class.
Previous work shows (e.g. in [21]), that both for activity recognition and for intensity estimation, accelerometers outperform gyroscopes. Therefore, from all \(3\) IMUs, only data from the accelerometers is used in the subsequent data processing steps.
Recently, new error metrics were introduced for continuous activity recognition, e.g. insertion, merge, overfill [39, 41]. However, contrary to activity recognition in home or industrial settings, for physical activity monitoring, the frame by frame metrics (precision, recall, F-measure and accuracy: all derivable from the confusion matrix) are sufficient, as discussed in [28].
References
Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz a M, Strath SJ, O’Brien WL, Bassett DR, Schmitz KH, Emplaincourt PO, Jacobs DR, Leon a S (2000) Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 32(9):498–504
Alimoglu F, Alpaydin E (1996) Methods of combining multiple classifiers based on different representations for ben-based handwritten digit recognition. In Proceedings of 5th Turkish Artificial Intelligence and Artificial Neural Networks Symposium (TAINN), Istanbul, Turkey
Bache K, Lichman M (2013) UCI machine learning repository. University of California, Irvine, School of Information and Computer Sciences, 2013. http://archive.ics.uci.edu/ml
Berchtold M, Budde M, Gordon D, Schmidtke H, Beigl M (2010) ActiServ: activity recognition service for mobile phones. In Proceedings of IEEE 14th International Symposium on Wearable Computers (ISWC), Seoul, South Korea
Eibl G, Pfeiffer KP (2002) How to make AdaBoost. M1 work for weak base classifiers by changing only one line of the code. In Proceedings of 13th European Conference on Machine Learning (ECML), Helsinki, Finland, pp 72–83
Ermes M, Pärkkä J, Cluitmans L (2008) Advancing from offline to online activity recognition with wearable sensors. In Proceedings of 30th Annual International IEEE EMBS Conference, Vancouver, BC, Canada, pp 4451–4454
Faddoul JB, Chidlovskii B, Gilleron R, Torre F (2012) Learning multiple tasks with boosted decision trees. In Proceedings of 2012 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD), Bristol, UK, pp 681–696
Fisher RA (1936) The use of multiple measurements in taxonomic problems. Ann Eugen 7:179–188
Freund Y (1995) The strength of weak learnability. Inf Comput 121(2):256–285
Freund Y, Schapire RE (1997) A desicion-theoretic generalization of on-line learning and an application to boosting. J Comput Syst Sci 55(1):119–139
Frey PW, Slate DJ (1991) Letter recognition using Holland-style adaptive classifiers. Mach Learn 6(2):161–182
Friedman J, Hastie T, Tibshirani R (2000) Additive logistic regression: a statistical view of boosting. Ann Stat 28(2):337–407
Gómez-Verdejo V, Ortega-Moral M, Arenas-García J, Figueiras-Vidal AR (2006) Boosting by weighting critical and erroneous samples. Neurocomputing 69(7–9):679–685
Haskell WL, Lee I-M, Pate RR, Powell KE, Blair SN, Franklin BA, Macera CA, Heath GW, Thompson PD, Bauman A (2007) Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc 39(8):1423–1434
Huang J, Ertekin S, Song Y, Zha H, Giles CL (2007) Efficient multiclass boosting classification with active learning. In Proceedings of SIAM International Conference on Data Mining (SDM), Minneapolis, MN, USA
Jin X, Hou X, Liu C-L (2010) Multi-class AdaBoost with hypothesis margin. In Proceedings of 20th International Conference on Pattern Recognition (ICPR), Washington, DC, USA, pp 65–68
Kuncheva LI (2004) Combining pattern classifiers: methods and algorithms. Wiley-Interscience, Hoboken
Lee M-H, Kim J, Kim K, Lee I, Jee SH, Yoo SK (2009) Physical activity recognition using a single tri-axis accelerometer. In Proceedings of World Congress on Engineering and Computer Science (WCECS), San Francisco, CA, USA
Long X, Yin B, Aarts RM (2009) Single-accelerometer based daily physical activity classification. In Proceedings of 31st Annual International IEEE EMBS Conference, Minneapolis, MN, USA, pp 6107–6110
Mease D, Wyner A (2008) Evidence contrary to the statistical view of boosting. J Mach Learn Res 9:131–156
Pärkkä J, Ermes M, Antila K, van Gils M, Mänttäri A, Nieminen H (2007) Estimating intensity of physical activity: a comparison of wearable accelerometer and gyro sensors and 3 sensor locations. In Proceedings of 29th Annual International IEEE EMBS Conference, Lyon, France, pp 1511–1514
Pärkkä J, Cluitmans L, Ermes M (2010) Personalization algorithm for real-time activity recognition using PDA, wireless motion bands, and binary decision tree. IEEE Trans Inf Technol Biomed 14(5):1211–1215
Quinlan JR, Compton PJ, Horn KA, Lazarus L (1986) Inductive knowledge acquisition: a case study. In Proceedings of 2nd Australian Conference on Applications of Expert Systems, Sydney, Australia, pp 137–156
Quinlan JR (1993) C4.5: programs for machine learning. Morgan Kaufmann, San Mateo
Quinlan JR (1996) Bagging, boosting and C4.5. In Proceedings of 13th National Conference on Artificial Intelligence (AAAI), Portland, OR, USA, pp 725–730
Ravi N, Dandekar N, Mysore P, Littman M (2005) Activity recognition from accelerometer data. In Proceedings of 17th Conference on Innovative Applications of Artificial Intelligence (IAAI), Pittsburgh, PA, USA, pp 1541–1546
Reiss A, Weber M, Stricker D (2011) Exploring and extending the boundaries of physical activity recognition. In: Proceedings of 2011 IEEE International Conference on Systems., Man and Cybernetics (SMC), Workshop on Robust Machine Learning Techniques for Human Activity Recognition, Anchorage, AK, USA, pp 46–50
Reiss A, Stricker D (2012) Creating and benchmarking a new dataset for physical activity monitoring. In Proceedings of 5th Workshop on Affect and Behaviour Related Assistance (ABRA), Crete, Greece
Reiss A, Stricker D (2012) Introducing a new benchmarked dataset for activity monitoring. In Proceedings of IEEE 16th International Symposium on Wearable Computing (ISWC), Newcastle, UK, pp 108–109
Reiss A, Stricker D (2013) Aerobic activity monitoring: towards a long-term approach. International Journal of Universal Access in the Information Society (UAIS)
Reiss A, Hendeby G, Stricker D (2013) Towards robust activity recognition for everyday life: methods and evaluation. In Proceedings of 7th International Conference on Pervasive Computing Technologies for Healthcare (PervasiveHealth), Venice, Italy
Reiss A, Hendeby G, Stricker D (2013) Confidence-based multiclass AdaBoost for physical activity monitoring. In Proceedings of IEEE 17th International Symposium on Wearable Computing (ISWC), Zurich, Switzerland, pp 13–20
Reiss A, Stricker D (2013) Personalized mobile physical activity recognition. In Proceedings of IEEE 17th International Symposium on Wearable Computing (ISWC), Zurich, Switzerland, pp 25–28
Schapire RE (1990) The strength of weak learnability. Mach Learn 5(2):197–227
Schapire RE (1997) Using output codes to boost multiclass learning problems. In Proceedings of 14th International Conference on Machine Learning (ICML), Nashville, TN, USA, pp 313–321
Schapire RE, Singer Y (1999) Improved boosting algorithms using confidence-rated predictions. Mach Learn 37(3):297–336
Siebert JP (1987) Vehicle recognition using rule based methods. Turing Institute, Glasgow, Scotland
Tapia EM, Intille SS, Haskell W, Larson K, Wright J, King A, Friedman R (2007) Real-time recognition of physical activities and their intensities using wireless accelerometers and a heart rate monitor. In Proceedings of IEEE 11th International Symposium on Wearable Computing (ISWC), Boston, MA, USA, pp 1–4
van Kasteren T, Alemdar H, Ersoy C (2011) Effective performance metrics for evaluating activity recognition methods. In Proceedings of 24th International Conference on Architecture of Computing Systems (ARCS), Como, Italy
Vezhnevets A, Vezhnevets V (2005) Modest AdaBoost - teaching AdaBoost to generalize better. In Proceedings of 15th International Conference on Computer Graphics and Applications (Graphicon), Novosibirsk, Russia
Ward JA, Lukowicz P, Gellersen HW (2011) Performance metrics for activity recognition. ACM Trans Intell Syst Technol 2(1):Article No. 6
Wu X, Kumar V, Quinlan JR, Ghosh J, Yang Q, Motoda H, McLachlan GJ, Ng A, Liu B, Yu PS, Zhou Z, Steinbach M, Hand DJ, Steinber D (2007) Top 10 algorithms in data mining. Knowl Inf Syst 14(1):1–37
Zhang T, Yu B (2005) Boosting with early stopping: convergence and consistency. Ann Stat 33(4):1538–1579
Zhao Z, Chen Y, Liu J, Shen Z, Liu M (2011) Cross-people mobile-phone based activity recognition. In Proceedings of 22nd International Joint Conference on Artificial Intelligence (IJCAI), Barcelona, Spain, pp 2545–2550
Zhu J, Rosset S, Zou H, Hastie T (2005) Multi-class Adaboost. Technical Report 430, Department of Statistics, University of Michigan
Zhu J, Zou H, Rosset S, Hastie T (2009) Multi-class Adaboost. Stat Interface 2:349–360
Acknowledgments
The work of Attila Reiss was partially supported by the collaborative project SimpleSkin under contract with the European Commission (#323849) in the FP7 FET Open framework. The support is gratefully acknowledged.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Reiss, A., Hendeby, G. & Stricker, D. A novel confidence-based multiclass boosting algorithm for mobile physical activity monitoring. Pers Ubiquit Comput 19, 105–121 (2015). https://doi.org/10.1007/s00779-014-0816-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00779-014-0816-x