离散傅里叶变换DFT

最新推荐文章于 2024-06-23 14:07:20 发布

wuxianfeng1987

最新推荐文章于 2024-06-23 14:07:20 发布

阅读量573

点赞数

分类专栏：机器学习&深度学习文章标签：快速傅里叶变换

本文链接：https://blog.csdn.net/wuxianfeng1987/article/details/55834465

版权

机器学习&深度学习专栏收录该内容

20 篇文章 1 订阅

订阅专栏

离散傅里叶变换的公式为：

Xk=∑n=0N−1x(n)ωknN ${X_k} = \sum\limits_{n = 0}^{N - 1} {x(n)\omega _N^{kn}}$

ωN=e−j2πN ${\omega _N} = {e^{ - j\frac{{2\pi }}{N}}}$

分成奇数列和偶数列后：

(0)

Xk=∑n=0N2−1x2nω2knN+∑n=0N2−1x2n+1ω(2n+1)kN ${X_k} = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _N^{2kn}} + \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _N^{(2n + 1)k}}$
(1)

ω 2 n k N = (e - j 2 π N) (2 n) k = e - j 2 π N 2 n k = ω n k N 2

$\begin{array}{l} \omega _N^{2nk} = {({e^{ - j\frac{{2\pi }}{N}}})^{(2n)k}}\\ = {e^{ - j\frac{{2\pi }}{{\frac{N}{2}}}nk}}\\ = \omega _{\frac{N}{2}}^{nk} \end{array}$

由（0）（1）得：

(2)

X k = \sum n = 0 N 2 - 1 x 2 n ω 2 k n N + \sum n = 0 N 2 - 1 x 2 n + 1 ω (2 n + 1) k N = \sum n = 0 N 2 - 1 x 2 n ω 2 k n N + ω k N * \sum n = 0 N 2 - 1 x 2 n + 1 ω 2 n k N = \sum n = 0 N 2 - 1 x 2 n ω k n N 2 + ω k N \sum n = 0 N 2 - 1 x 2 n + 1 ω k n N 2

$\begin{array}{l} {X_k} = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _N^{2kn}} + \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _N^{(2n + 1)k}} \\ = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _N^{2kn}} + \omega _N^k*\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _N^{2nk}} \\ = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _{\frac{N}{2}}^{kn}} + \omega _N^k\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _{\frac{N}{2}}^{kn}} \end{array}$

因为

ωknN2 $\omega _{\frac{N}{2}}^{kn}$ 的周期是

N2 $\frac{N}{2}$ ，所以 :

ωknN2=ω(k+N2)nN2 $\omega _{\frac{N}{2}}^{kn} = \omega _{\frac{N}{2}}^{(k + \frac{N}{2})n}$ （3）

而

ωkN $\omega _N^k$ 的周期是N，

ωk+N2N=−ωkN $\omega _N^{k + \frac{N}{2}} = - \omega _N^k$ （4）

由（2）（3）（4）得:

Xk+N2=∑n=0N2−1x2nωknN2+ωkN∑n=0N2−1x2n+1ωknN2 ${X_{k + \frac{N}{2}}} = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _{\frac{N}{2}}^{kn}} + \omega _N^k\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _{\frac{N}{2}}^{kn}}$ {k=0—-N/2-1}（5.1）

Xk+N2=∑n=0N2−1x2nωknN2−ωkN∑n=0N2−1x2n+1ωknN2 ${X_{k + \frac{N}{2}}} = \sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _{\frac{N}{2}}^{kn}} - \omega _N^k\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n + 1}}\omega _{\frac{N}{2}}^{kn}}$ { k=N—-N-1}（5.2）

如何是8点的FFT

这里写图片描述

∑n=0N2−1x2nωknN2 $\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _{\frac{N}{2}}^{kn}}$ 可以看成是偶数项的DFT，同上有：

∑n=0N2−1x2nωknN2=∑n=0N4−1x2(2n)ωk(2n)N2+∑n=0N4−1x2(2n+1)ωk(2n+1)N2 $\sum\limits_{n = 0}^{\frac{N}{2} - 1} {{x_{2n}}\omega _{\frac{N}{2}}^{kn}} = \sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{2(2n)}}\omega _{\frac{N}{2}}^{k(2n)}} + \sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{2(2n + 1)}}\omega _{\frac{N}{2}}^{k(2n + 1)}} \\$

=∑n=0N4−1x4nωknN4+ωkN2∑n=0N4−1x4n+2ωknN4 $= \sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n}}\omega _{\frac{N}{4}}^{kn}} + \omega _{\frac{N}{2}}^k\sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n + 2}}\omega _{\frac{N}{4}}^{kn}}$

X1(k)=∑n=0N4−1x4nωknN4+ωkN2∑n=0N4−1x4n+2ωknN4 ${X_1}(k) = \sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n}}\omega _{\frac{N}{4}}^{kn}} + \omega _{\frac{N}{2}}^k\sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n + 2}}\omega _{\frac{N}{4}}^{kn}}$

X1(k+N4)=∑n=0N4−1x4nωknN4−ωkN2∑n=0N4−1x4n+2ωknN4 ${X_1}(k + \frac{N}{4}) = \sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n}}\omega _{\frac{N}{4}}^{kn}} - \omega _{\frac{N}{2}}^k\sum\limits_{n = 0}^{\frac{N}{4} - 1} {{x_{4n + 2}}\omega _{\frac{N}{4}}^{kn}}$
这里写图片描述

wuxianfeng1987

关注

0
点赞
踩
1

收藏

觉得还不错? 一键收藏
0
评论
离散傅里叶变换DFT

离散傅里叶变换的公式为： Xk=∑n=0N−1x(n)ωknN{X_k} = \sum\limits_{n = 0}^{N - 1} {x(n)\omega _N^{kn}} ωN=e−j2πN{\omega _N} = {e^{ - j\frac{{2\pi }}{N}}} 分成奇数列和偶数列后： (0) Xk=∑n=0N2−1x2nω2knN+∑n=0N2−1x2n+1ω
复制链接

扫一扫

专栏目录