ABSTRACT
The advancement of the internet to the paradigm of the Internet
of Things (IoT) has brought to society new ways of generating,
sharing and using information. The evolution of computing capacity
and energy savings in IoT equipament combined with better
software can enabled several new applications, among which we
can highlight the monitoring of people’s health through pervasive
devices connected to the body. In view of this, this work proposes
an algorithm to detect atypical situations such as falls in the elderly
and other groups that need health care using accelerometers
contained in wearable devices, particularly smartwatches. For the
experimental evaluation of the proposed algorithm, a database that
contains data from wearable sensors, environmental sensors, and
visual devices was employed. The metrics used in the evaluation
were accuracy, precision, recall and f1-score, with recall being the
most relevant metric in the context. Results show that the best
configuration of the algorithm is able to identify falls with 96%
recall and F1-score of 90%.