Research On Tracking Performance Of RLS Algorithm And The Stationarity Of Time Series

Speech signal analysis is the basis of speech signal processing, it is the premise to realize the application such as communication of voice, speech recognition and speech synthesis. The Speech signal analysis throughout the speech signal processing and application of has quite an important position. Speech signal is time-varying signal which has a "short-term stability", so any for speech signal research and analysis are based on “the short-term analysis".In speech signal processing, clear and dullness voiced judgment is an important part of speech signals processing, It is the key factor in the follow-up work,such as gene cycle detection, speech synthesis and so on. This paper research on tracking performance of RLS algorithm and time series stationarity base on the clear and dullness voiced judgment of 26 English letters’ audio files.At present, there are many kinds of methods of clear and dullness voiced judgment, such as clear and dullness voiced judgment based on analysis of quantitative recursion, clear and dullness voiced judgment based on phase space reconstruction and so on. Howerver, the characteristics of these algorithms is that the accuracy of the judgment is relatively low. These methods judge clear and dullness voiced by using threshold and their choice of parameters can also affect the effect of judgment.Aimed at drawing these disadvantages of the algorithms mentioned above, this paper proposes two new algorithms of clear and dullness voiced judgment,named clear and dullness voiced judgment based on Robust Recurrence Least Square algorithm and clear and dullness voiced judgment based on Recurrence Least Square Lattice algorithm. These algorithms are designed with the idea of one algorithm, named clear and dullness voiced judgment based on Recurrence Least Square algorithm, which was proposed by TianYing Wu. Then, this paper explores the respective optimal order-number and the optimal frame-length of these algorithms, and compares the time performance, Tracking accuracy and effect of judgment of above three algorithms. Experiments show that the algorithms proposed in this paper are effective and feasible. Furthermore, this paper analyzes the correlation between each algorithm and its according order-number and length of frame.In the end, this paper also proposes a algorithm of based on the stationarity of timeseries, studies the influence of time window on stationarity, and compares the effect of judgment between this algorithm and the three algorithms mentioned above. Similarily, experiments show that the method proposed in this paper is effective and feasible.