Optimal selection of mother wavelet for classifying human activities from acceleration signals

Falls are major problems that could have happened to elderly, and could cause paralysis, hip fractures, disabilities or accidental deaths. A human activity classification from acceleration signals may be an important process in fall prevention or detection. An algorithm which combines the wavelet transform and the multilayer perceptron neural network is an effective tool for classifying complicate signals. In order to optimize the classification, this paper aims to compare the performances of the algorithm which uses different mother wavelets. In our experiments, 5 volunteers who were healthy with the ages between 21 to 25 year old were asked to attach a tri-axial accelerometer at the right side of their waists. Next, the volunteers were asked to perform 5 daily-life activities: 1) walking, 2) standing up from a chair, 3) sitting down on a chair, 4) lying down on a bed, and 5) getting up from a bed; and 5 falling events: 1) forward falling, 2) backward falling, 3) falling to the right side, 4) falling to the left side, and 5) falling when standing up. In this paper, there are 2 experiments. In the first experiment, the algorithm was used to classify the real activity of the acceleration signals. Then the output of the algorithm can be any activity from ten activities. In the second experiment, the algorithm was used to detect the falling events. Then the output of the algorithm has 2 values; the falling event or the daily-life activity. The mother wavelets which an; used to evaluate the performances of the classification were Daubechies, Coiflet, Symlet, and Biorthogonal. From the experiments of the both of the human activity classification and the falling detection, the algorithm which used the Biorthogonal mother wavelet showed the best performances.