A Hybrid DWT–PCA–SVM Framework for Smartphone-Based Human Activity Recognition in Medical Monitoring Applications: The Ghanaian Context

The increasing prevalence of chronic diseases and the growing need for remote healthcare delivery in Ghana highlight the importance of efficient, low-cost, and intelligent patient monitoring systems. This study presents a hybrid Discrete Wavelet Transform–Principal Component Analysis–Support Vector Machine (DWT–PCA–SVM) framework for smartphone-based Human Activity Recognition (HAR), designed to support real-time medical monitoring applications within the Ghanaian context. The proposed framework leverages the Discrete Wavelet Transform (DWT) to denoise raw accelerometer signals and extract discriminative time–frequency features. Principal Component Analysis (PCA) is then employed to reduce feature dimensionality while retaining over 95% of the total variance, thereby optimizing computational efficiency for mobile devices. Finally, a Support Vector Machine (SVM) classifier with a fine-tuned radial basis function (RBF) kernel is used to accurately distinguish between common daily activities such as walking, sitting, standing, lying down, and jogging. Experimental evaluations using the UCI HAR benchmark dataset demonstrated that the hybrid DWT–PCA–SVM model achieved superior recognition performance, with an average accuracy of 66.0%, surpassing the baseline SVM by approximately 6%. Additionally, wavelet-based preprocessing enhanced signal quality with an average SNR improvement of 3.5 dB and RMSE reduction of 30% compared to conventional Butterworth filtering. In the Ghanaian healthcare context, the proposed approach provides a practical foundation for mobile health (mHealth) solutions capable of continuous, non-invasive activity monitoring using widely available smartphones. By enabling early detection of abnormal activity patterns and promoting patient self-management, this framework supports the national agenda for digital healthcare transformation and offers a scalable model for resource-limited environments across sub-Saharan Africa. Keywords: Human Activity Recognition, DWT, PCA, SVM, Chronic Disease, Ghana