Early detection of high fall risk is an essential component of fall prevention in older adults. Wearable sensors can provide valuable insight into daily-life activities; biomechanical features extracted from such inertial data have been shown to be of added value for the assessment of fall risk. Body-worn sensors such as accelerometers can provide valuable insight into fall risk. Currently, biomechanical features derived from accelerometer data are used for the assessment of fall risk. Here, we studied whether deep learning methods from machine learning are suited to automatically derive features from raw accelerometer data that assess fall risk. We used an existing dataset of 296 older adults. We compared the performance of three deep learning model architectures (convolutional neural network (CNN), long short-term memory (LSTM) and a combination of these two (ConvLSTM)) to each other and to a baseline model with biomechanical features on the same dataset. The results show that the deep learning models in a single-task learning mode are strong in recognition of identity of the subject, but that these models only slightly outperform the baseline method on fall risk assessment. When using multi-task learning, with gender and age as auxiliary tasks, deep learning models perform better. We also found that preprocessing of the data resulted in the best performance (AUC = 0.75). We conclude that deep learning models, and in particular multi-task learning, effectively assess fall risk on the basis of wearable sensor data.

Many algorithms on health monitoring from ambient sensor networks assume that only a single person is present in the home. We present an unsupervised method that models visit behaviour. A Markov modulated multidimensional non-homogeneous Poisson process (M3P2) is described that allows us to model weekly and daily variations and to combine multiple data streams, namely the front-door sensor transitions and the general sensor transitions. The results from nine months of sensor data collected in the apartment of an elderly person show that our model outperforms the standard Markov modulated Poisson process (MMPP).

No abstract

No abstract

This paper presents a study of sensor data from a person who developed Alzheimer's disease during a 4-year monitoring period and who is monitored with simple ambient sensors in her home. Our aim is to find data analysis methods that reveal relevant changes in the sensor pattern that occur before the diagnosis. We focus on the quantification of regularity, which is identified as a relevant indicator for the assessment of a disease such as Alzheimer's. Two unsupervised methods are studied. Restricted Boltzmann Machines are trained and the resulting weights are visualized to see whether there are changes in regularity in the behavioral pattern. Fast Fourier Transformation is applied to the sensor data and the spectral characteristics are determined and compared with the same purpose. Both methods reveal changes in the pattern between different periods. Both methods therefore are useful in quantifying and understanding changes in the regularity of the daily pattern.