Recently, Human Activity Recognition (HAR) has gained meaningful information for a human being. A wearable sensor like an accelerometer, small and simple to perform, has opened the room for scientists to explore an initial understanding of ubiquitous computing. The wearable sensor has begun to receive attention among researchers in some respects to conduct their studies in a wide area of recognition of human activity. Recent ADL discusses not only simple activities but also cater to the broad categories of complex activities. However, when involving enormous numbers of a subject, the accuracy of recognition tends to reduce. Although a different subject performed the same activity, the acceleration signal acquired considerably differs. This is due to the distinct pattern of action for each subject based on various factors such as subject age, gender, emotion and personality. Thus, by enhancing the accuracy of recognition of ADL, this article proposes the framework for addressing the subject independent matter. The signal acquired from an accelerometer sensor will undergo a segmentation process to extract important features. Some of the characteristics may be meaningless in some instances to determine the class. Therefore, proposing a variety of features to select the most relevant features that can lead to accuracy above 90%. Also, this article outlined a brief empirical evaluation of previous related work. Using several machine learning algorithms, this preliminary work will be examined and analyzed.

Quantitative R2* map is an important liver disease indicator. However, the availability of R2* map is limited by the long scan time. In this study, we present a new paradigm to predict R2* and B0 maps from dual echo images. A self-attention deep convolutional neural network is trained and validated, where promising accuracy has been obtained. The proposed quantitative parametric mapping approach has a potential to eliminate the necessity for additional data acquisition other than clinical routine.