Characteristics of accelerometer signals to classify events as falls or activities of daily living
School Name
Spring Valley High School
Grade Level
10th Grade
Presentation Topic
Math and Computer Science
Presentation Type
Mentored
Abstract
In the present era, falls are one of the most prominent causes for distress among the elderly. If the occurrence of a fall is not recognized the elder is left lying immobile, suffering from their injuries; This time of immobility is known as the long lie, and most fall related deaths occur within it. It is proposed that all falls cause similar structural vibrations when they occur and thus can produce signals with similar traits. Furthermore, it is proposed that other events (known as Activities of Daily Living or ADLs) produce vibrations that can be categorized as their own, a theory which will be measured by specificity, a percentage representative of the number of correctly identified ADLs. To test these hypotheses, structural vibrations have been recorded with an accelerometer and were analysed for alpha, amplitude and duration values. These values were run through a classification code that used thresholds and limits to identify each event as a fall or an ADL. It was concluded from the variation of classification by method that a combination of all three previously mentioned parameters will result in the highest specificity. The data were analysed through a two sample T test and it was determined that the occurrence of false alarms being highest for events where a bag was dropped was statistically significant: t(4)=3.747, p=0.018899. These results do not support the original hypothesis but do show the potential for use of this system with more precise calibration.
Recommended Citation
Patterson, Elaine R., "Characteristics of accelerometer signals to classify events as falls or activities of daily living" (2015). South Carolina Junior Academy of Science. 22.
https://scholarexchange.furman.edu/scjas/2015/all/22
Start Date
4-11-2015 9:45 AM
End Date
4-11-2015 10:00 AM
Characteristics of accelerometer signals to classify events as falls or activities of daily living
In the present era, falls are one of the most prominent causes for distress among the elderly. If the occurrence of a fall is not recognized the elder is left lying immobile, suffering from their injuries; This time of immobility is known as the long lie, and most fall related deaths occur within it. It is proposed that all falls cause similar structural vibrations when they occur and thus can produce signals with similar traits. Furthermore, it is proposed that other events (known as Activities of Daily Living or ADLs) produce vibrations that can be categorized as their own, a theory which will be measured by specificity, a percentage representative of the number of correctly identified ADLs. To test these hypotheses, structural vibrations have been recorded with an accelerometer and were analysed for alpha, amplitude and duration values. These values were run through a classification code that used thresholds and limits to identify each event as a fall or an ADL. It was concluded from the variation of classification by method that a combination of all three previously mentioned parameters will result in the highest specificity. The data were analysed through a two sample T test and it was determined that the occurrence of false alarms being highest for events where a bag was dropped was statistically significant: t(4)=3.747, p=0.018899. These results do not support the original hypothesis but do show the potential for use of this system with more precise calibration.
