Feature Engineering for Human Activity Recognition

Human activity recognition (HAR) techniques can significantly contribute to the enhancement of health and life care systems, in particular, for elderly people. These techniques, which generally employ measured collected data from wearable sensors or those embedded in smartphones, have therefore attracted a lot of interest recently. However, low sampling rate, noisy recorded signals and relatively few participating subjects make the recognition task of a group of activities more challenging. In this paper, a random forest-based classifier for HAR is proposed. The classifier is trained using a set of time-domain features extracted from raw sensor data segmented into 5-seconds window with 50% overlapping. To evaluate the performance of the proposed classifier compared to other machine learning techniques such as support vector machines and Artificial Neural Networks, several simulation experiments are conducted on the benchmark wearable human activity recognition folder (WHARF) accelerometer dataset. The proposed model shows high recognition rates for the 12 activities in WHARF dataset, e.g. 94.6% for brushing teeth, 94.1% for descending stairs, 90.7% for drinking, and 92.7% for using telephone. With regard to implementation issuses of processing time and model size on disk, the proposed recognition system is shown to be efficient from practical point of view and is expected to be suitable for the implementation in hand-held devices such as smartphones with their limited memory and computational resources. Sensitivity analysis of random forest model parameters shows that the optimal number of estimators is 100 while the maximum depth for the random forest is 20 achieving a precision of 86.2%. Also, t-test is employed to reveal the rank of features in terms of their importance. Using this test, the size of feature vector is reduced from 3 7 to 24 achieving a precision of 82.2%. The overall average accuracy of the proposed classifier of 84.86% is superior to the current state-of-the-art rate of 79.13%.

[1]  Y.-K. Lee,et al.  Human Activity Recognition via an Accelerometer-Enabled-Smartphone Using Kernel Discriminant Analysis , 2010, 2010 5th International Conference on Future Information Technology.

[2]  Liping Jin,et al.  Human Activity Recognition from Sensor-Based Large-Scale Continuous Monitoring of Parkinson’s Disease Patients , 2017, 2017 IEEE/ACM International Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE).

[3]  Leandro Bezerra Marinho,et al.  A New Approach to Human Activity Recognition Using Machine Learning Techniques , 2016, ISDA.

[4]  Hanyu Wang,et al.  LSTM-CNN Architecture for Human Activity Recognition , 2020, IEEE Access.

[5]  Parisa Rashidi,et al.  Human Activity Recognition Using Inertial, Physiological and Environmental Sensors: A Comprehensive Survey , 2020, IEEE Access.

[6]  Ciprian Dobre,et al.  Human Physical Activity Recognition Using Smartphone Sensors , 2019, Sensors.

[7]  Hang Su,et al.  A Smartphone-Based Adaptive Recognition and Real-Time Monitoring System for Human Activities , 2020, IEEE Transactions on Human-Machine Systems.

[8]  Alan F. Smeaton,et al.  Optimised Convolutional Neural Networks for Heart Rate Estimation and Human Activity Recognition in Wrist Worn Sensing Applications , 2020, ArXiv.

[9]  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).

[10]  William Robson Schwartz,et al.  Novel approaches to human activity recognition based on accelerometer data , 2018, Signal Image Video Process..

[11]  Guilin Chen,et al.  Human Activity Recognition in a Smart Home Environment with Stacked Denoising Autoencoders , 2016, WAIM Workshops.

[12]  Eunji Kim,et al.  Interpretable and Accurate Convolutional Neural Networks for Human Activity Recognition , 2020, IEEE Transactions on Industrial Informatics.

[13]  Javier Villamizar,et al.  Human activity recognition using improved complete ensemble EMD with adaptive noise and long short-term memory neural networks , 2020, Biocybernetics and Biomedical Engineering.

[14]  P. Aguirre.  Feature Enrichment in Human Activity Recognition , 2018 .

[15]  A. M. Khan,et al.  Accelerometer signal-based human activity recognition using augmented autoregressive model coefficients and artificial neural nets , 2008, 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[16]  Jozsef Suto,et al.  Human activity recognition using neural networks , 2014, Proceedings of the 2014 15th International Carpathian Control Conference (ICCC).

[17]  Fadi Al Machot,et al.  A review on applications of activity recognition systems with regard to performance and evaluation , 2016, Int. J. Distributed Sens. Networks.

[18]  Tae-Seong Kim,et al.  A single tri-axial accelerometer-based real-time personal life log system capable of human activity recognition and exercise information generation , 2011, Personal and Ubiquitous Computing.

[19]  João Gama,et al.  Human Activity Recognition Using Inertial Sensors in a Smartphone: An Overview , 2019, Sensors.

[20]  Haoxi Zhang,et al.  A Novel IoT-Perceptive Human Activity Recognition (HAR) Approach Using Multihead Convolutional Attention , 2020, IEEE Internet of Things Journal.

[21]  Tae-Seong Kim,et al.  A Triaxial Accelerometer-Based Physical-Activity Recognition via Augmented-Signal Features and a Hierarchical Recognizer , 2010, IEEE Transactions on Information Technology in Biomedicine.

[22]  E. D. Lemaire,et al.  Evaluation of a smartphone human activity recognition application with able-bodied and stroke participants , 2016, Journal of NeuroEngineering and Rehabilitation.

[23]  Matthias Struck,et al.  User-friendly system for recognition of activities with an accelerometer , 2010, 2010 4th International Conference on Pervasive Computing Technologies for Healthcare.

[24]  A. Zakaria,et al.  Activity recognition using accelerometer sensor and machine learning classifiers , 2018, 2018 IEEE 14th International Colloquium on Signal Processing & Its Applications (CSPA).

[25]  Xi Long,et al.  Single-accelerometer-based daily physical activity classification , 2009, 2009 Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[26]  Thenkurussi Kesavadas,et al.  Using mini minimum jerk model for human activity classification in home-based monitoring , 2015, 2015 IEEE International Conference on Rehabilitation Robotics (ICORR).

[27]  Wilhelmiina Hämäläinen,et al.  Jerk-based feature extraction for robust activity recognition from acceleration data , 2011, 2011 11th International Conference on Intelligent Systems Design and Applications.

[28]  Davide Anguita,et al.  A Public Domain Dataset for Human Activity Recognition using Smartphones , 2013, ESANN.

[29]  Ronald Poppe,et al.  A survey on vision-based human action recognition , 2010, Image Vis. Comput..

[30]  Magda B. Fayek,et al.  An enhanced method for human action recognition , 2015, Journal of advanced research.

[31]  Sung-Bae Cho,et al.  Human activity recognition with smartphone sensors using deep learning neural networks , 2016, Expert Syst. Appl..

[32]  Fulvio Mastrogiovanni,et al.  A public domain dataset for ADL recognition using wrist-placed accelerometers , 2014, The 23rd IEEE International Symposium on Robot and Human Interactive Communication.

[33]  Faicel Chamroukhi,et al.  Physical Human Activity Recognition Using Wearable Sensors , 2015, Sensors.

[34]  Priscila L. R. Aguirre,et al.  Autoregressive Modeling of Wrist Attitude for Feature Enrichment in Human Activity Recognition , 2018 .

[35]  Abdulhamit Subasi,et al.  Human activity recognition using machine learning methods in a smart healthcare environment , 2020 .

[36]  Nishant Doshi,et al.  Human Activity Recognition: A Survey , 2019, Procedia Computer Science.

[37]  Macarena Espinilla,et al.  Sensor-Based Datasets for Human Activity Recognition – A Systematic Review of Literature , 2018, IEEE Access.