Speech character analysis based on non-linear method

The current speech character analysis method is based on frequency feature, Regular periodical signal and dis-ordered signal may have the same spectrum but their complexity is different, therefore, to some extent, frequency spectrum lost some information, according to our experiment complexity can remedy this. The traditional complexity measures, including the KC and C1,C2 complexity, have the coarse granulation effect produced in symbolizing the time series, and lost very much detail information which reflect speech character sometimes. In this paper, we advanced one new complexity measure -Partition Measure Complexity can reserve much more details, and can reflect more exact and comprehensive character of the time series. We apply, this complexity measure to analyze speech character, and obtained analysis outcome.