Abstract:
Prehensile movement is crucial for activities of daily living (ADLs) such as grooming
and self-care. In humans, the hand is the primary device utilized for prehensile movements.
Recognizing the relationship between a hand's prehensile patterns and accelerometer data
can be instrumental in developing assisted and enhanced functional hand devices. An
accelerometer measures the linear acceleration acting on the part of the body where the
sensor is placed. This thesis demonstrates the usefulness of several features, based on accelerometer data, towards recognizing prehensile movement of the hand while performing
47 movements, grips and neutral rest. Particular emphasis is given to measuring the use-
fulness of the features to identify movement transitions. A random forest classifier is used
to recognize motion onset and offset as well as various phases of movement. The results
showed that the accelerometer-based features were effective in recognizing motion onset
and offset and moments of transitions. However, they were not as effective in recognizing
various phases of prehensile movement.
