This book introduces the readers to a comprehensive idea to implement machine learning-based physical activity recognition frameworks. This book covers the challenges and their respective solutions of machine learning-based human activity monitoring and recognition frameworks.

A novel feature selection method, modified guided regularized random forest, is introduced to accurately select the most relevant and important features to address the “curse-of-dimensionality” and “overfitting” issues. Ensemble learning, Random projection-based ELM, feature fusion, and deep learning frameworks with attention mechanisms are explored for human activity recognition in the rest of the chapters.

The importance of transitional activities is also discussed concerning hemiplegia gait analysis and the concept of online change point detection segmentation method is also introduced. Finally, the book ends with a flexible activity recognition and real-time monitoring system (Flexi-HAMR), which can efficiently monitor and recognize activities using online, real-time data streams and also update the model dynamically for any new activity such as Parkinsonian gait for early disease prediction.

. .