Josh's Review — 数字信号处理
Part 2 离散傅里叶级数 DFS

1. 傅里叶变换的几种形式

1.1 非周期连续时间信号的傅里叶变换

  非周期连续时间信号的傅里叶变换(连续时间傅里叶变换,Continuous-Time Fourier Transform, CTFT)的形式为:

\[\begin{equation} X_a\left(j\Omega\right)=\int_{-\infty}^{\infty}{x_a\left(t\right)e^{-j\Omega t}\mathrm{d}t} \end{equation}\]

\[\begin{equation} x_a\left(t\right)=\frac{1}{2\pi}\int_{-\infty}^{\infty}{X_a\left(j\Omega\right)e^{j\Omega t}\mathrm{d}\Omega} \end{equation}\]

1.2 周期连续时间信号的傅里叶变换

  周期连续时间信号的傅里叶变换(连续傅里叶级数,Continous Fourier Series, CFS)的形式为:

\[\begin{equation} X\left(m\Omega\right)=\frac{1}{t_p}\int_{-\frac{t_p}{2}}^{\frac{t_p}{2}}{x_a\left(t\right)e^{-jm\Omega t}\mathrm{d}t} \end{equation}\]

\[\begin{equation} x_a\left(t\right)=\sum_{m=-\infty}^{\infty}{X\left(m\Omega\right)e^{jm\Omega t}} \end{equation}\]

其中 \(t_p\) 为信号的周期,频域相邻谱线间角频率增量 \(\Omega=\dfrac{2\pi}{t_p}\)

1.3 非周期离散时间信号的傅里叶变换

  非周期离散时间信号的傅里叶变换(离散时间傅里叶变换,Discrete-Time Fourier Transform, DTFT)的形式为:

\[\begin{equation} X\left(e^{j\omega}\right)=\sum_{n=-\infty}^{\infty}{x\left(n\right)e^{-j\omega n}} \end{equation}\]

\[\begin{equation} x\left(n\right)=\frac{1}{2\pi}\int_{-\pi}^{\pi}{X\left(e^{j\omega}\right)e^{j\omega n}\mathrm{d}\omega} \end{equation}\]

取样频率 \(f_s=\dfrac{1}{T}\),取样角频率 \(\Omega_s=\dfrac{2\pi}{T}\),取样数字频率\(\omega_s=2\pi\)

1.4 周期离散时间信号的傅里叶变换

  周期离散时间信号的傅里叶变换(离散傅里叶级数,Discrete Fourier Series, DFS)的形式为:

\[\begin{equation} \widetilde{X}\left(k\right)=\left.\widetilde{X}\left(e^{j\omega}\right)\right|_{\omega=k\frac{2\pi}{N}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{kn}} \end{equation}\]

\[\begin{equation} \widetilde{x}\left(n\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}} \end{equation}\]

其中 \(N\) 为一个周期内的点数,时域相邻谱线间隔为 \(T\) ,频域相邻谱线间隔为 \(\dfrac{1}{NT}\),以数字频率表征时为 \(\dfrac{2\pi}{N}\)

由以上,在一个域中对函数进行取样,两取样点间增量的倒数,必是另一个域中的周期。

2. 离散傅里叶级数(DFS)的推导

2.1 正变换

  令 \(x\left(n\right)\)\(N\) 点长的离散序列,则其 DTFT(上标“~”用来表征周期性)

\[\begin{equation} \widetilde{X}\left(e^{j\omega}\right)=\sum_{n=-\infty}^{\infty}{x\left(n\right)e^{-j\omega n}}=\sum_{n=0}^{N-1}{x\left(n\right)e^{-j\omega n}} \end{equation}\]

\(\widetilde{X}\left(e^{j\omega}\right)\) 取样,使其称为周期离散频率函数,由此时域序列 \(x\left(n\right)\) 周期化为 \(\widetilde{x}\left(n\right)\),令其谱线间隔为 \(T\),则时域序列的周期为 \(NT\),所以对频谱取样的谱间距为 \(\dfrac{1}{NT}\),以数字频率表征时,谱间距

\[\begin{equation} \omega_\mathrm{I}=\frac{2\pi}{N} \end{equation}\]

因此,以数字频率 \(\omega\) 为变量的 \(X\left(e^{j\omega}\right)\) 被离散化时,其变量 \(\omega\) 变为

\[\begin{equation} \omega=k\omega_\mathrm{I}=k\frac{2\pi}{N} \end{equation}\]

因此离散周期序列 \(\widetilde{x}\left(n\right)\) 的傅里叶级数可写为

\[\begin{equation} \widetilde{X}\left(k\right)=\left.\widetilde{X}\left(e^{j\omega}\right)\right|_{\omega=k\frac{2\pi}{N}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{kn}} \end{equation}\]

并将数字域频率用 \(k\) 简化表示。且

\[\begin{equation} \widetilde{X}\left(k+N\right)=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}\left(k+N\right)n}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j2\pi n}e^{-j\frac{2\pi}{N}kn}=}\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}=}\widetilde{X}\left(k\right) \end{equation}\]

\(\widetilde{X}\left(k\right)\) 是以 \(N\) 为周期的周期函数。

2.2 逆变换

  将正变换两边乘以 \(e^{j\frac{2\pi}{N}kr}\),并在一个周期内求和,有

\[\begin{equation} \sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kr}}=\sum_{k=0}^{N-1}{\left(\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}\right)e^{j\frac{2\pi}{N}kr}}=N\left[\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)\left(\frac{1}{N}\sum_{k=0}^{N-1}e^{j\frac{2\pi}{N}k\left(r-n\right)}\right)}\right] \end{equation}\]

由正交定理

\[\begin{equation} \frac{1}{N}\sum_{k=0}^{N-1}e^{j\frac{2\pi}{N}k\left(r-n\right)}= \begin{aligned} \begin{cases} 1,&r=n\\ 0,&r≠n \end{cases} \end{aligned} \end{equation}\]

\[\begin{equation} \sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kr}}=N\left.\left[\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)}\right]\right|_{n=r}=N\widetilde{x}\left(r\right) \end{equation}\]

可得逆离散傅里叶变换

\[\begin{equation} \widetilde{x}\left(n\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}} \end{equation}\]

\[\begin{equation} \widetilde{x}\left(n+mN\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}k\left(n+mN\right)}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}e^{j2\pi km}}=\widetilde{x}\left(n\right) \end{equation}\]

\(\widetilde{x}\left(n\right)\) 是以 \(N\) 为周期的周期函数。

3. 离散傅里叶级数(DFS)的性质

3.1 线性性质

\[\begin{equation} \mathrm{DFS}\left[a{\widetilde{x}}_1\left(n\right)+{b\widetilde{x}}_2\left(n\right)\right]=a{\widetilde{X}}_1\left(k\right)+b{\widetilde{X}}_2\left(k\right) \end{equation}\]

3.2 时移性质

\[\begin{equation} \mathrm{DFS}\left[\widetilde{x}\left(n+m\right)\right]=W_N^{-mk}\widetilde{X}\left(k\right) \end{equation}\]

证明:

\[\begin{equation} \begin{aligned} \mathrm{DFS}\left[\widetilde{x}\left(n+m\right)\right]&=\sum_{n=0}^{N-1}{\widetilde{x}\left(n+m\right)W_N^{nk}}=\sum_{n=m}^{N-1+m}{\widetilde{x}\left(n\right)W_N^{\left(n-m\right)k}}=\sum_{n=m}^{N-1+m}{\widetilde{x}\left(n\right)W_N^{nk}W_N^{-mk}} \\ &\xlongequal{利用\widetilde{x}\left(n\right)和W_N^{nk}的周期性}\left(\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{nk}}\right)W_N^{-mk}=W_N^{-mk}\widetilde{X}\left(k\right) \end{aligned} \end{equation}\]

3.3 频移性质

\[\begin{equation} \mathrm{IDFS}\left[\widetilde{X}\left(k+l\right)\right]=W_N^{nl}\widetilde{x}\left(n\right) \end{equation}\]

证明:

\[\begin{equation} \begin{aligned} \mathrm{IDFS}\left[\widetilde{X}\left(k+l\right)\right]&=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k+l\right)W_N^{-kn}}=\frac{1}{N}\sum_{k=l}^{N-1+l}{\widetilde{X}\left(k\right)W_N^{-\left(k-l\right)n}}=\frac{1}{N}\sum_{k=l}^{N-1+l}{\widetilde{X}\left(k\right)W_N^{-kn}W_N^{nl}} \\ &\xlongequal{利用\widetilde{X}\left(n\right)和W_N^{-nk}的周期性}\frac{1}{N}\left(\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}}\right)W_N^{nl}=W_N^{nl}\widetilde{x}\left(n\right) \end{aligned} \end{equation}\]

3.4 时域周期卷积

\[\begin{equation} {\widetilde{x}}_1\left(n\right)\odot{\widetilde{x}}_2\left(n\right)=\mathrm{IDFS}\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right] \end{equation}\]

证明:

\[\begin{equation} \begin{aligned} \mathrm{IDFS}\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right]&=\frac{1}{N}\sum_{k=0}^{N-1}{\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right]W_N^{-nk}}\\ &=\frac{1}{N}\sum_{k=0}^{N-1}{\left[\sum_{m=0}^{N-1}{\widetilde{x}}_1\left(m\right)W_N^{mk}\right]{\widetilde{X}}_2\left(k\right)W_N^{-nk}}\\ &=\frac{1}{N}\sum_{k=0}^{N-1}\left[\sum_{m=0}^{N-1}{\widetilde{x}}_1\left(m\right){\widetilde{X}_2\left(k\right)W}_N^{-\left(n-m\right)k}\right]\\ &=\sum_{m=0}^{N-1}{\widetilde{x}}_1\left(m\right)\left[\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}_2\left(k\right)W}_N^{-\left(n-m\right)k}\right]\\ &=\sum_{m=0}^{N-1}{\widetilde{x}}_1\left(m\right){\widetilde{x}}_2\left(n-m\right) \\ &={\widetilde{x}}_1\left(n\right)\odot{\widetilde{x}}_2\left(n\right) \end{aligned} \end{equation}\]

注:此处用 \(\odot\) 表示圆周卷积,不同教材中圆周卷积的符号表述可能不同。

3.5 频域周期卷积

\[\begin{equation} \mathrm{DFS}\left[{\widetilde{x}}_1\left(n\right){\widetilde{x}}_2\left(n\right)\right]=\frac{1}{N}\left[{\widetilde{X}}_1\left(k\right)\odot{\widetilde{X}}_2\left(k\right)\right] \end{equation}\]

证明:

\[\begin{equation} \begin{aligned} \mathrm{DFS}\left[{\widetilde{x}}_1\left(n\right){\widetilde{x}}_2\left(n\right)\right]&=\sum_{n=0}^{N-1}{\widetilde{X}_1\left(n\right){\widetilde{x}}_2\left(n\right)W_N^{kn}}=\sum_{n=0}^{N-1}\left[\frac{1}{N}\sum_{r=0}^{N-1}{\widetilde{X}_1\left(r\right)W_N^{-rn}}\right]{\widetilde{x}}_2\left(n\right)W_N^{kn}\\ &=\frac{1}{N}\sum_{n=0}^{N-1}\left[\sum_{r=0}^{N-1}{\widetilde{X}_1\left(r\right){\widetilde{x}}_2\left(n\right)W_N^{\left(k-r\right)n}}\right]\\ &=\frac{1}{N}\sum_{r=0}^{N-1}{\widetilde{X}}_1\left(r\right)\left[\sum_{k=0}^{N-1}{\widetilde{X}_2\left(n\right)W_N^{\left(k-r\right)n}}\right]\\ &=\frac{1}{N}\sum_{r=0}^{N-1}{\widetilde{X}}_1\left(r\right){\widetilde{X}}_2\left(k-r\right)\\ &=\frac{1}{N}\left[{\widetilde{X}}_1\left(k\right)\odot{\widetilde{X}}_2\left(k\right)\right] \end{aligned} \end{equation}\]

注:此处用 \(\odot\) 表示圆周卷积,不同教材中圆周卷积的符号表述可能不同。

参考文献

王世一. 数字信号处理, 修订版. 北京理工大学出版社, 1997.