Change point detection and estimation of change times are important signal processing aspects in patient monitoring, rehabilitation supervision and biomedical research. In this paper, a sequential algorithm is presented for computerized detection of multiple changes in human motor responses based on the Maximum Likelihood principle. Unlike traditional threshold-based approaches which usually assume a step-like change profile the algorithm uses a ramp-step model instead. This allows to consider gradual changes as well which are typical for the dynamics of energy-limited signals like, e.g., force and position recordings. The algorithm is successfully applied to position signals measured in tapping experiments.
展开▼
