A Probabilistic Fuzzy Classifier for Motion Intent Recognition

Human motion intentions are commonly identified from signals collected by sensors. However, these signals are susceptible to various noises and uncertainties, leading to unreliable identification accuracy. To enhance robustness and reliability, this study proposes a probabilistic fuzzy classifier for motion intent recognition in the presence of noise. The method begins by analyzing and estimating the stochastic effect of disturbances on the kernel parameter and regularization parameter. Subsequently, a new objective function is formulated, incorporating the distribution of these parameters. To solve this objective function, a probabilistic inference strategy is developed to estimate the distribution. Using this distribution information, a solving strategy for the fuzzy model is devised. By constructing the distribution relationship between disturbances and model parameters and incorporating probabilistic information in the model, the proposed method demonstrates enhanced robustness and reliability for motion intention recognition under noisy conditions. Experimental results validate that the proposed algorithm improves accuracy and robustness for motion intention recognition and outperforms other state-of-the-art recognition algorithms in noise environments.