Robust spatio-temporal descriptors for real-time SVM-based fall detection

We propose a SVM-based approach to detect falls in several home environments using an optimised descriptor adapted to real-time tasks. We build an optimised spatio-temporal descriptor named STHFa_SBFS using several combinations of transformations of geometrical features, thanks to feature selection. We study the combinations of usual transformations of the features (Fourier Transform, Wavelet transform, first and second derivatives). Automatic feature selection allows to show that the best tradeoff between classification performance and processing time is obtained combining the original low-level features with their first derivative. Hence, we evaluate the robustness of the fall detection regarding location changes. We propose a realistic and pragmatic protocol which enables performance to be improved by updating the training in the current location with normal activities records. An embedded implementation of the fall detection based on a smart camera prototype is briefly depicted and demonstrates that a compact version of the detector can be deployed.