Arm Movements Effects in Response to Posture Instability

In recent years, because of an increasing aging population there are higher incidences of falling according to epidemiological reports. Because of this high frequency the prevention of falls becomes a major concern. Evidence of the high occurrence and significant cost of falls on health-related quality of life, significant financial load on the health care system, and on their social impact has been provided by various epidemiological studies. Falls are the second leading cause of traumatic brain injury (TBI), which is a major cause of death in many countries, especially the United States. Balance impairments are frequent and particularly high among people who suffer from stroke, TBI, incomplete spinal cord injuries, Parkinson’s disease, multiple sclerosis and diabetic peripheral neuropathy, and in general for people who suffer from different neurological disorders. For all of these groups, balance disorders have a major social and quality of life implications, which require attention and exploration of effective ways to evaluate risk and develop training programs that prevent falls. According to the literature, the most important factors for fall prevention are suitable training programs and the availability of feasible and cost-effective comprehensive risk measurement [1, 2]. This thesis describes the acquisition of acceleration data of a human body while maintaining balance on a balance board with three-axis accelerometers. Three different algorithms of balance region detection, the wavelet transform, and the neural network were developed to segment and classify the unstable regions of the accelerometer signal. To simplify the calculation of these algorithms vector processing technique was used. The experimental results show that arms have an effective role in the improvement of balance. From the balance region detection the duration and amount of activity can be found which will be good for prediction of falls. The wavelet transform is the best way to separate unstable periods from one another. For classification of stable and unstable parts of movements, the neural network is the best technique. It is effective to compare the amount of stable and unstable parts in more detail. The results suggest the specific role of the dominant and non-dominant arms.