Abstract: According to the time-varying speech signal, a novel method combining LLE (locally linear embedding) with fuzzy kernel clustering algorithm was proposed for speech visualization, where LLE could reduce the nonlinear dimensionality of the speech features and then the fuzzy kernel clustering algorithm was used for clustering analysis, i.e. the Mercer kernel function was used to change the data in original space into a high-dimensional eigenspace through nonlinear mapping, and then the fuzzy clustering analysis was made in the high-dimensional eigenspace. Thus, after the kernel function mapping, the original inherent features of speech were highlighted to improve the position-based speech visualization. 720 data in Chinese vowels were obtained from 10 male and 10 female students' speech in lab, the results of simulation experiments show the feasibility and validity of the method.