In this study, a number of different fall scenarios and the
possible changes in accelerometer data were identified. When
the system is initiated, the data continuously generated by the
accelerometer is tracked. An accelerometer reads data at a
frequency of 13 – 15 times per second, and the last 20
differences between the accelerometer readings are stored in
the watch memory. Every time the accelerometer data is read,
the last 20 differences are checked to determine if more than 2
values are above the predetermined threshold value. In
addition, it detects sudden changes in the 3 dimensional
orientation of the watch, from the x, y and z values provided by
the accelerometer. A detection of sudden changes in two axes
(exceeding threshold value) that potentially match fall patterns
previously identified causes, the axis data to be sent to the main
computer. Figure 7 shows a sample fall data that is generated
by an accelerometer