Based Multi-frequency Analysis Of Speech Signals Ridge Extraction Algorithm

Language as the main way of communication with each other, occupies a pivotal position in people’s daily life. Speech signal processing as a branch of phonetics in signal processing has an enormous role for us to understand the phonetics and prove the daily lives of people in communication.For speech signal processing, three parameters are run through the voice signal processing, which always been:instantaneous frequency, instantaneous amplitude and instantaneous phase. In theory, if we know the three parameters, we can get the most of the characteristic parameters of speech signal, and the signal for more in-depth research. So how to obtain these three parameters, how to make these three parameters relate to each other, which become a main problem of the speech signal we study..The time-frequency analysis method is one of the main methods for the time-frequency signal processing. We can obtain the information of signal in this method. According to the classification of signal, we can divide the method into two more detailed methods which are linear time-frequency analysis method and nonlinear frequency analysis method. The former method aims at a linear system or stationary signals and the latter method is mainly for non-linear systems or non-stationary signals.The work of the paper is mainly summarized as follows:Firstly, the paper researched the method of Time-frequency analysis. This paper mainly researched two major methods,which are Hilbert-Huang Transform time-frequency analysis method and Wavelet Transform time-frequency analysis method. This paper elaborated the definition, principles and the nature of the above methods.Secondly, for the two above nonlinear time-frequency analysis, we researched corresponding practical algorithm separately. Empirical Mode Decomposition (EMD) method is for Hilbert-Huang Transform time-frequency analysis method and Wavelet Ridge Extraction Algorithm is for Wavelet time-frequency analysis method. For each algorithm, we researched the detail from the reason why the algorithm was proposed to the implementation steps and then some examples were listed. The validity of the algorithms were tested in these examples. Additionally, the paper focused on improving the ridge of wavelet transform extraction algorithm and analyzed performance of the improved algorithm.Then, time-frequency analysis was introduced in the speech signal. Three methods were applied in the processing of speech signal, which are the spectrogram analysis method, empirical mode decomposition method and wavelet ridge extraction method modulus maxima value method. advantages and disadvantages of the various algorithms were obtained by comparing the results of the three methods.Finally, we summarized the Multi-ridge extraction algorithm.