Skip to Main Content
This study is part of the ongoing care assessment platform project, which involves monitoring vital signs and daily activity profile information of chronic disease patients undergoing cardiac rehabilitation. In this study, we have focussed on detecting walking activity from a cardiac rehab session which includes many other high intensity activities such as biking and rowing, using waist worn accelerometers. Walking is an important measure useful to assess the mobility of elderly people. Various methods have been proposed in the literature to identify walking from waist worn accelerometer signals based on wavelet, frequency and computational intelligence methods. Wavelet based approach, due to its feasibility to be implemented in real time with low computational complexity, good accuracies and also the ability to provide good time frequency resolution, has been the most desirable approach. In this study, we have evaluated and compared six wavelet decomposition based measures to detect walk from other high intensity activities. The different measures were derived using anterior-posterior, vertical, medio-lateral and signal vector magnitude (SVM) acceleration signals. The results show that all these measures can discriminate walking from other high intensity activities and the SVM based measure was the most efficient (89.14% sensitivity and 89.97 % specificity).