FALL DETECTOR WITH RANDOM CLASSIFIER OPTIMIZATION

Fall accidents are one of the main causes of injuries and accidental deaths for elder people. Automated fall detectors can be helpful for detecting the accident and sending an emergency call. For this purpose, a wearable fall detector device was developed, which served to collect a training set and to test certain classifiers. The current approach uses a random search algorithm to optimize a given model for discrimination of fall events from activities of daily living by measuring tri-axial accelerometer data. In the current state the device is able to recognize significant falls and can trigger an emergency call over mobile networks.