摘要
语音识别广泛应用于人机交互、安全识别等相关领域,语音信号分类是语音识别的重要基础。语音信号分类主要借助混沌特性的相关特征对语音信号进行研究。目前,语音信号分类相关研究主要有模型训练分类和特征提取两种方法。模型训练分类法需要大量数据的支撑,而且训练过程复杂、训练时间长。特征提取法需要提取大量不同特征进行分析,过程复杂。文中在特征提取法的基础上提出一种基于李雅普诺夫指数的语音信号混沌特性分类方法。该方法以混沌理论中相空间重构为基础,分别采用互信息法求取延迟时间、Cao方法求取嵌入维数、小数据量法求最大李雅普诺夫指数,然后探究各类语音信号的分布特点,并对其进行分类。
Speech recognition is widely applied in human-machine interaction,security recognition and other related fields. The classification of speech signal is an important basis for speech recognition and it is mainly based on the relevant characteristics of chaotic characteristics to study speech signal. At present,the related researches of speech signal classification mainly include model training classification and feature extraction. The former needs a lot of data with complex training process and long training time. The latter needs to extract a large number of different features for analysis,which is also complex in process. In this paper,based on the feature extraction method,we propose a chaotic speech signal classification method based on Lyapunov index. On the basis of phase space reconstruction in chaotic theory,we respectively calculate the delay time by mutual information method,the embedded dimension by Cao method and the maximum Lyapunov index by small-data volume method,then explore the distribution characteristics of various speech signals and classify them.
引文
[1]朱琦,酆广增,肖海勇.基于模式识别的语音分类方法[J].南京邮电学院学报:自然科学版,2000,20(4):29-33.
[2] GAO Y,SHAO S,XIAO X,et al. Using pseudo amino acid composition to predict protein subcellular location:approached w ith Lyapunov index,Bessel function,and Chebyshev filter[J]. Amino Acids,2005,28(4):373-376.
[3] ELSNER J B,TSONIS A A. Phase space reconstruction[M]//Singular spectrum analysis. US:Springer,1996:143-155.
[4] SUZUKI H. Takens’embeddingtheorem[J]. Journal of Japan Society for Fuzzy Theory&Systems,1998,10:82-86.
[5]张淑清,贾健,高敏,等.混沌时间序列重构相空间参数选取研究[J].物理学报,2010,59(3):1576-1582.
[6]吕小青,曹彪,曾敏,等.确定延迟时间互信息法的一种算法[J].计算物理,2006,23(2):184-188.
[7] CAO Liangyue. Practical method for determining the minimum embedding dimension of a scalar time series[J]. Physica D:Nonlinear Phenomena,1997,110(1-2):43-50.
[8] SU Y,LIANG S,ZENG C,et al. Study on nonlinear variable selection based on false nearest neighbours in KPLS subspace[J]. International Journal of Advancements in Computing Technology,2012,4(18):324-332.
[9] ROSENSTEIN M T,COLLINS J J,DELUCA C J. A practical method for calculating largest Lyapunov exponents from small data sets[J]. Physica D:Nonlinear Phenomena,1993,65(1-2):117-134.
[10]张勇,陈天麒,陈滨.计算最大Lyapunov指数的推广小数据量法[J].电子科技大学学报,2004,33(3):254-257.
[11]鲁铁定,陶本藻,周世健.基于整体最小二乘法的线性回归建模和解法[J].武汉大学学报:信息科学版,2008,33(5):504-507.
[12]王庆福.汉语语音的局部线性预测及其编码应用[D].南京:南京大学,2004.
[13]焦伟华,席晓革.英语发音与单词音标拼读[M].郑州:河南大学出版社,2011.
[14]叶龙.综合自然拼读法与国际音标构建英语拼读拼写方案的研究设计[D].长沙:湖南大学,2013.