Wavelet transform based fall detection using wearable accelerometers

Abstract

Falls are identified as a major health risk, especially for the elderly people and are considered a major obstacle to independent living. Quick medical response is desired in case of a fall event. However, the fall may leave the elderly person in such a state that the elderly may be unable to call for help on his/her own. Automatic and fast detection of falls would decrease the health risks associated with the falls and would make independent living safer for the elderly people. In this thesis, we propose an automatic fall detection system that uses a wearable accelerometer and incorporates wavelet transform as a feature extraction method. We conducted experiments to investigate the performance of the system under the effect of several factors including fall properties, selection of wavelet transform parameters and sensor platform types. Results indicate that our proposed approach is robust with high fall detection performance. The fall detection mechanism was realized using the wearable sensors that were part of an indoor monitoring environment, namely WeCare. WeCare not only provided the necessary sensing capabilities for the fall detection but it also made available several communication and notification methods. Using these methods, we were able to notify caregivers in case of fall detection. In this thesis, we also describe the WeCare system and the integration of our fall detection study into it.