基于时变自回归模型的音频和语音信号分析毕业论文中英文资料对照外文翻译文献综述.doc

PAGE \* MERGEFORMAT PAGE \* MERGEFORMAT 1 PAGE \* MERGEFORMAT PAGE \* MERGEFORMAT 1 PAGE \* MERGEFORMAT 12 PAGE \* MERGEFORMAT 13 文中英文资料对照外文翻译文献综述 附 录 附录A 外文翻译—原文部分 Time Varying autoregressive modeling of Audio and speech signals Aki H?rm?1 , Marko Juntunen2 1.Helsinki University of Technology, Laboratory of Acoustics and Audio Signal Processing P.O.Box 3000 FIN-02015, Espoo, FINLAND, e-mail: Aki.Harma@hut.fi 2.Department of Applied Physics, University of Kuopio P.O.Box 1627, FIN-70211, Kuopio, FINLAND, e-mail: Jari.Kaipio@uku.fi Conventional linear predictive techniques for modeling of speech and audio signals are based on an assumption that a signal is stationary within each analysis frame. However,natural signals are often continuously timevarying, i.e., nonstationary.Therefore this assumption might not be well justified.In this paper, we study a timevarying autoregressive(TVAR) modeling technique in which this restriction is relaxed.A frequency-warped formulation of the Subba RaoLiporace TVAR algorithm is introduced in the article. Theapplicability of the presented methodology to various speechand audio signal processing tasks is illustrated and discussed.It is also shown that the TVAR scheme yields an efficientparametrization for timevarying sounds. 1 Introduction Linear prediction, LP, is a standard technique in speech and audio coding [1] and in several other applications for analysis and synthesis of audio and speech signals. Conventional LP techn- iques utilize framebased autoregressive spectral modeling,e.g., autocorrelation or autocovariance method of LP.In conventional LP, it is assumed that a signal x(t) can beexpressed as a linear com bination of the previous samples by （1） Here, ak are a set of fixed set of p coefficients which can be estimated from a signal frame of T samples, and e(t) is a prediction error, or a residual, signal. In warped LP [2], this is modified by replacing x(t?k) in (1) by dk[x(t)]