英语论文网

1. Introduction
The wavelet transform involves joint time-frequency representation with locality and, thus, it is wx useful in non-stationary signal analysis 1–3 . The wavelet transform manifests as a hyperbolic time-frequency tiling, for example, in a spectrogram rep-wx resentation 2 . In particular, high frequency analysis is adapted to narrow time-width while low frequency analysis uses wider time-width. The adaptive prop-erty of a standard wavelet transform is fixed, i.e., given a scaling function it accommodates all applica-tions in identical manner. In other words, a standard
wavelet transform, except for its inherent time-frequency adaptability, is unable to adapt itself to specific application.

1 引言

小波变换是局部联合时频表示和，因此，它是有用的在非平稳信号分析Wx 1–3。小波变换的表现为双曲线的时频块，例如，在图2代表WX表示。特别是，高频率的分析适用于狭窄的时间宽度，而低频分析使用广泛的时间宽度。一个标准的小波变换是固定的，即自适应特性，给出了尺度函数可容纳在相同的方式，所有的应用。换句话说，一个标准的小波变换，除了其固有频率的适应性，不能适应特定的应用。

An irregular tree decomposition is expected to yield better performance than a standard wavelet tree wx decomposition 4–7 . Accordingly, an embedded zero-tree based on dyadic wavelet decomposition has wx already been proposed to prune a wavelet tree 4 . In this paper, we explore identifying an arbitrary sub-band decomposition tree by splitting the wavelets into wavelet packets, offering improved adaptability to specific applications.

一个不规则的树分解的预期收益率比标准的小波树分解4–Wx 7更好的性能。因此，嵌入式零树基于二进小波分解具有蜡质已经提出的小波树的修剪4。在本文中，我们探讨通过将小波为小波包识别一个任意子带分解树，提供改进的适应性的具体应用

A wavelet transform can be described by its wavelet orthonormal basis generated from a mother wx wavelet by dilations and translations 1,3,8 . The orthonormal wavelet basis typically provides a framework for studying images simultaneously at different levels. Using a multiresolution analysis, the successive ‘fine-to-coarse’ approximations can be represented by a nested sequence of subspaces while the information differences between two resolution
levels are described by a sequence of complement wx subspaces 9 . Multiresolution analysis can be con-sidered as a recursive filtering using a pair of wavelet filters and, eventually, by the scaling function as wx determined by two-level equations 1,3,8 . While the sequence of nested subspaces associated with a mul-tiresolution analysis is described by dilations and translations of a scaling function, the sequence of complement subspaces is described by dilations and wx translations of a mother wavelet 8,9 . A wavelet function can be considered as a high-pass signal. It has bandpass characteristic since fre-quency localization requires that a high-pass signal decays rapidly with increasing frequency. The nar-rower the bandpass bandwidth, the better the fre-quency resolution. The spectral support of a dyadic wavelet is a dyadic interval whose size changes as 2j where j is the resolution level. However, the expo-nential behavior of dyadic support is fixed. For a higher-resolution wavelet, one makes use of the next higher resolution level. Alternatively, in this paper, we refine the wavelet frequency resolution at a level
without going to the next level. We demonstrate the advantages of wavelet splitting over a standard wavelet transform in terms of compression and re-covery scores. Since wavelet splitting uses the sa本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。