Video Based Mobility Monitoring of Elderly People Using Deep Learning Models

In recent years, the number of older people living alone has increased rapidly. Innovativevision systems to remotely assess people's mobility can help healthy, active, and happy aging. In the relatedliterature, the mobility assessment of older people is not yet widespread in clinical practice. In addition,the poor availability of data typically forces the analyses to binary classification, e.g. normal/anomalousbehavior, instead of processing exhaustive medical protocols. In this paper, real videos of elderly peopleperforming three mobility tests of a clinical protocol are automatically categorized, emulating the complexevaluation process of expert physiotherapists. Videos acquired using low-cost cameras are initiallyprocessed to obtain skeletal information. A proper data augmentation technique is then used to enlargethe dataset variability. Thus, significant features are extracted to generate a set of inputs in the form oftime series. Four deep neural network architectures with feedback connections, even aided by a preliminaryconvolutional layer, are proposed to label the input features in discrete classes or to estimate a continuousmobility score as the result of a regression task. The best results are achieved by the proposed Conv-BiLSTMclassifier, which achieves the best accuracy, ranging between 88.12% and 90%. Further comparisons withshallow learning classifiers still prove the superiority of the deep Conv-BiLSTM classifier in assessingpeople's mobility, since deep networks can evaluate the quality of test executions.