Chapter16 Evaluation of a Time Reversal Method with Dynamic Time Warping Matching Function for Human Fall Detection Using Structural Vibrations Ramin Madarshahian, Juan M. Caicedo, and Diego Arocha Zambrana Abstract Falls are one of the predominant concerns of the elderly living at home. Commercial systems, such as wearable pendants that are pressed in an emergency, provide a viable solution when a fall occurs. However, wearable systems have a low compliance, especially in patients with diseases such as Alzheimer’s or other forms of dementia. Monitoring changes in the environment provides the possibility of reducing the compliance challenges for those patients. Computer vision techniques is an example of environmental monitoring. However, some patients might be concerned about their privacy when having cameras in their homes. Monitoring the vibrations of the patient’s dwelling is another alternative. Classification of the acceleration recorded signals becomes important to determine if a fall has occurred. This paper proposes the use of the Time Reversal Method (TRM) with Dynamic Time Warping (DTW) for classifying structural accelerations produced by different human actions. The potential classification is studied by releasing objects at different heights. A statistical study is performed to determine the importance of different factors to the application of the proposed technique. These factors are distance to the sensor, type of object used to impact the floor and intensity of the impact. Results indicate that the technique is most sensitive to the type of object, indicating the potential for human fall detection. Results also show interaction between the height in which the object was released and the type of object. Distance between the location of impact and the sensor is not an important factor but has an effect on the standard deviation of results. Keywords Fall detection • Factorial design • Dynamic time warping (DTW) • Structural vibration • Classification 16.1 Introduction Falls is one of the main reasons for hospitalization due to accidents in elder populations [1–3]. Furthermore, serious falls are usually not quickly reported to medical personnel or caregivers because the patient is unable to call for help [4]. Fall detection systems can reduce the time between the incident and the arrival time of medical attention. Fall detection systems can be classified in wearable, vision based, and ambient systems [5]. Wearable sensor, which are directly attached to the body of patient have been found reliable for fall detection, but difficult to wear in some situations, especially while taking showers [6, 7]. Moreover, some patients such as Alzheimer patients might forget or decide not to wear the devices [8]. Computer vision techniques solves the compliance problem but raises privacy concerns. Furthermore, falls should happen in the visual range of the camera to be detected. The use of accelerometers to monitor floor vibrations can be considered as an alternative ambient fall system. Tracking vibration on the patient’s dwelling, have the potential to solve both compliance challenges and privacy concerns [9–12]. Vibration-based methods are feasible because accelerometers are not expensive and installation is easy. Previous work performed at the Structural Dynamics and Intelligent Infrastructure (SDII) laboratory at the University of South Carolina and by other research groups shows that events are easily detectible [9, 13]. However, little work has been done in the classification of the events. This paper investigates what factors are important for event classification. The Dynamic Time Warping (DTW) is used for event classification for comparing a potential event with a series of reference signals. An experimental test is designed to study the effect of several parameters in the signal classification. These are: R. Madarshahian ( ) • J.M. Caicedo • D.A. Zambrana University South Carolina, 300 Main St., Columbia, SC 29208, USA e-mail: mdrshhn@email.sc.edu; caicedo@cec.sc.edu; arochazd@email.sc.edu H.S. Atamturktur et al. (eds.), Model Validation and Uncertainty Quantification, Volume 3: Proceedings of the 32nd IMAC, A Conference and Exposition on Structural Dynamics, 2014, Conference Proceedings of the Society for Experimental Mechanics Series, DOI 10.1007/978-3-319-04552-8__16, © The Society for Experimental Mechanics, Inc. 2014 171
RkJQdWJsaXNoZXIy MTMzNzEzMQ==