On-Board Feature Extraction from Acceleration Data for Activity Recognition

Modern wearable devices are equipped with increasingly powerful microcontrollers and therefore are increasingly capable of doing computationally heavy operations, such as feature extraction from sensor data. This paper quantifies the time and energy costs required for on-board computation of features on acceleration data, the reduction achieved in subsequent communication load, and the impact on daily activity recognition in terms of classification accuracy. The results show that platforms based on modern 32-bit ARM Cortex-M microcontrollers significantly benefit from on-board extraction of time-domain features. On the other hand, efficiency gains from computation of frequency domain features at the moment largely remain out of their reach.

[1]  Ming Tan,et al.  Cost-sensitive learning of classification knowledge and its applications in robotics , 2004, Machine Learning.

[2]  Tomasz Wiktorski,et al.  Enrichment of Machine Learning Based Activity Classification in Smart Homes Using Ensemble Learning , 2016, 2016 IEEE/ACM 9th International Conference on Utility and Cloud Computing (UCC).

[3]  Lei Liu,et al.  Human Daily Activity Recognition for Healthcare Using Wearable and Visual Sensing Data , 2016, 2016 IEEE International Conference on Healthcare Informatics (ICHI).

[4]  JeongGil Ko,et al.  Wireless Sensor Networks for Healthcare , 2010, Proceedings of the IEEE.

[5]  Qiang Wang,et al.  Low Power or High Performance? A Tradeoff Whose Time Has Come (and Nearly Gone) , 2012, EWSN.

[6]  Kuncup Iswandy,et al.  Feature selection with acquisition cost for optimizing sensor system design , 2006 .

[7]  Robert J. Piechocki,et al.  Designing Wearable Sensing Platforms for Healthcare in a Residential Environment , 2017, EAI Endorsed Trans. Pervasive Health Technol..

[8]  Gert R. G. Lanckriet,et al.  Recognizing Detailed Human Context in the Wild from Smartphones and Smartwatches , 2016, IEEE Pervasive Computing.

[9]  Mikkel Baun Kjærgaard,et al.  Smart Devices are Different: Assessing and MitigatingMobile Sensing Heterogeneities for Activity Recognition , 2015, SenSys.

[10]  Niall Twomey,et al.  The SPHERE Challenge: Activity Recognition with Multimodal Sensor Data , 2016, ArXiv.

[11]  R. Venkatesha Prasad,et al.  CoachMe: Activity Recognition using Wearable Devices for Human Augmentation , 2017, EWSN.

[12]  Qing Guo,et al.  Balancing energy, latency and accuracy for mobile sensor data classification , 2011, SenSys.

[13]  Ryan J. Halter,et al.  Amulet: An Energy-Efficient, Multi-Application Wearable Platform , 2016, SenSys.

[14]  Björn Eskofier,et al.  Approaching the accuracy–cost conflict in embedded classification system design , 2016, Pattern Analysis and Applications.

[15]  Paul J. M. Havinga,et al.  Activity Recognition Using Inertial Sensing for Healthcare, Wellbeing and Sports Applications: A Survey , 2010, ARCS Workshops.

[16]  Shan Lin,et al.  Toothbrushing Monitoring using Wrist Watch , 2016, SenSys.

[17]  Lior Rokach,et al.  The CASH algorithm-cost-sensitive attribute selection using histograms , 2013, Inf. Sci..

[18]  A. Kraskov,et al.  Estimating mutual information. , 2003, Physical review. E, Statistical, nonlinear, and soft matter physics.

[19]  Paul E. Hoffman,et al.  Concise Binary Object Representation (CBOR) , 2020, RFC.

[20]  Tina L Hurst,et al.  Physical activity classification using the GENEA wrist-worn accelerometer. , 2012, Medicine and science in sports and exercise.

[21]  Krisjanis Nesenbergs,et al.  Demo: Wearable Sensor System for Human Biomechanics Monitoring , 2016, EWSN.