A preliminary investigation of monitoring ADLs using wireless kinematic sensors

The objective of this on-going work is to evaluate the accuracy and reliability of wireless kinematic sensors in identifying basic activities of daily living (ADL). A preliminary trial was conducted consisting of 5 subjects; 3 male (mean: 23.6, SD: 2.41). Four kinematic sensors were placed on the subject; (a) mid sternum, (b) underneath the left armpit, (c) above the right hip and (d) the ankle of the dominant leg. A fifth sensor, the activPALtrade Trio Professional physical activity logger was used for comparison with the kinematic sensors. Each subject performed a range of basic activities´ in a controlled laboratory setting. Subjects were then asked to carry out similar self annotated activities in a random order and in an unsupervised environment. Feature sets of mean, standard deviation, frequency-domain entropy, discrete FFT coefficient and signal magnitude area are being calculated. These feature sets will be used to train several classifiers including decision tree´s, nearest neighbor, naive Bayes and support vector machines. Several meta-level classifiers will also be evaluated including boosting, bagging and plurality voting. We aim to identify the most reliable classifier and location for the kinematic sensor in indentifying basic ADLs.