Abstract

Using smartphone sensors for Human Activity Recognition (HAR) has become a crucial research field with applications in smart settings, fitness tracking, and healthcare. This work uses the widely used UCI HAR dataset to give a thorough comparative analysis of different machine learning and deep learning algorithms for HAR. Combining a deep convolutional neural network (CNN) architecture with six conventional machine learning algorithms—Random Forest, XGBoost, Support Vector Machines, k-Nearest Neighbors, and Logistic Regression— the results have been developed and assessed. To guarantee reliable performance evaluation, all models underwent a thorough evaluation process utilizing 5-fold stratified cross-validation. As our results show, the CNN architecture performed better than the others (96.2% accuracy), closely followed by the non-linear approach SVM (95.2%) and the linear method Logistic Regression (95.4%). The study provides valuable insights into the relative strengths of different algorithmic approaches for sensor-based activity recognition and offers practical guidance for selecting appropriate models for HAR applications.

References

• Anguita, D., et al. (2013). "A Public Domain Dataset for Human Activity Recognition Using Smartphones." ESANN
• Bulling, A., Blanke, U., & Schiele, B. (2014). A Tutorial on Human Activity Recognition Using Body-Worn Inertial Sensors. ACM Computing Surveys (CSUR), *46*(3), 1–33. https://doi.org/10.1145/2499621
• Reyes-Ortiz, J.-L., et al. (2016). "Transition-Aware Human Activity Recognition Using Smartphones." Neurocomputing
• Ronao, C. A., & Cho, S.-B. (2016). "Human activity recognition with smartphone sensors using deep learning neural networks." Expert Systems with Applications
• Hammerla, N. Y., et al. (2016). "Deep, Convolutional, and Recurrent Models for Human Activity Recognition using Wearables." IJCAI
• Bao L, Intille SS (2004) Activity recognition from user-annotated acceleration data. In: International Conference on Pervasive Computing. Springer, pp 1–17
• Wu W, Dasgupta S, Ramirez EE, Peterson C, Norman GJ (2012) Classification accuracies of physical activities using smartphone motion sensors. Journal of Medical Internet Research 14(5):e130
• Zhao Y, Li H, Wan S, Sekuboyina A, Hu X, Tetteh G, Piraud M, Menze B (2019) Knowledge-aided convolutional neural network for small organ segmentation. IEEE Journal of Biomedical and Health Informatics
• Ding S, Qu S, Xi Y, Wan S (2019) Stimulus-driven and concept-driven analysis for image caption generation, Neurocomputing. https://doi.org/10.1016/j.neucom.2019.04.095
• Xu X, Xue Y, Qi L, Yuan Y, Zhang X, Umer T, Wan S (2019) An edge computing-enabled computation offloading method with privacy preservation for internet of connected vehicles. Future Generation Computer Systems 96:89–100
• Li W, Liu X, Liu J, Chen P, Wan S, Cui X (2019) On improving the accuracy with auto-encoder on conjunctivitis. Applied Soft Computing, p 105489
• Jantawong P., Jitpattanakul A., Mekruksavanich S. Enhancement of Human Complex Activity Recognition using Wearable Sensors Data with InceptionTime Network; Proceedings of the 2021 2nd International Conference on Big Data Analytics and Practices (IBDAP); Bangkok, Thailand. 26–27August 2021; pp. 12–16.
• Challa S.K., Kumar A., Semwal V.B. A multibranch CNN-BiLSTM model for human activity recognition using wearable sensor data. Vis. Comput. 2022;38:4095–4109. doi: 10.1007/s00371-021-02283-3.
• Zaki, Z.; Shah, M.A.; Wakil, K.; Sher, F.: Logistic regression based human activities recognition. J. Mech. Continu. Math. Sci. 15(4),228–246 (2020)