Characterization Of Human Stability Using Vector Acceleration Signals

Biomedical signals carry information about a physiological event. The part of the signal pertaining to a specific event is called an epoch. Once the event has been determined, the corresponding waveform may be segmented and analyzed based on many parameters[1]. As falls have increased in recent years due to an aging population, it is important to gain insight to the reaction of an individual to perturbations. One common method of studying human reaction is by using a balance aperture. This thesis describes the physical actions that produce acceleration on a balance apparatus and captures the acceleration on an accelerometer. Algorithms were developed to segment the unstable periods of the accelerometer signal. Wavelets were used as well as non-linear filters. The non-linear filters increased the amplitudes of periods of instability, simple signal models of the output of the non-linear filters where formulated and analyzed. Vector processing techniques were also developed. The experimental results demonstrate that the acceleration during unstable periods can be differentiated by its frequency content, by its discontinuous nature and by using vector relationships. The algorithms were tested with five individuals and had over 80% accuracy.

Characterization Of Human Stability Using Vector Acceleration Signals

Biomedical signals carry information about a physiological event. The part of the signal pertaining to a specific event is called an epoch. Once the event has been determined, the corresponding waveform may be segmented and analyzed based on many parameters[1]. As falls have increased in recent years due to an aging population, it is important to gain insight to the reaction of an individual to perturbations. One common method of studying human reaction is by using a balance aperture. This thesis describes the physical actions that produce acceleration on a balance apparatus and captures the acceleration on an accelerometer. Algorithms were developed to segment the unstable periods of the accelerometer signal. Wavelets were used as well as non-linear filters. The non-linear filters increased the amplitudes of periods of instability, simple signal models of the output of the non-linear filters where formulated and analyzed. Vector processing techniques were also developed. The experimental results demonstrate that the acceleration during unstable periods can be differentiated by its frequency content, by its discontinuous nature and by using vector relationships. The algorithms were tested with five individuals and had over 80% accuracy.