The Discrete Fourier Transform (DFT)

Given a signal $x(\cdot)\in{\bf C}^N$, the spectrum is defined by^6.3

$$X(\omega_k) \isdef \left<x,s_k\right> = \sum_{n=0}^{N-1}x(n) \overline{s_k(n)}, \quad k=0,1,2,\ldots,N-1,$$

or, as is most often written

$$\zbox {X(\omega_k) \isdef \sum_{n=0}^{N-1}x(n) e^{-j\frac{2\pi k n}{N}}, \quad k=0,1,2,\ldots,N-1.}$$

That is, the $k$th sample $X(\omega_k)$ of the spectrum of $x$ is defined as the inner product of $x$ with the $k$th DFT sinusoid $s_k$. This definition is $N$ times the coefficient of projection of $x$ onto $s_k$, i.e.,

$$\frac{\left<x,s_k\right>}{\left\Vert\,s_k\,\right\Vert^2} = \frac{X(\omega_k)}{N}.$$

The projection of $x$ onto $s_k$ itself is

$${\bf P}_{s_k}(x) = \frac{X(\omega_k)}{N} s_k.$$

Since the $\{s_k\}$ are orthogonal and span ${\bf C}^N$, using the main result of the preceding chapter, we have that the inverse DFT is simply the sum of the projections:

$$x = \sum_{k=0}^{N-1}\frac{X(\omega_k)}{N} s_k$$

or, as we normally write,

$$\zbox {x(n) = \frac{1}{N} \sum_{k=0}^{N-1}X(\omega_k) e^{j\frac{2\pi k n}{N}}.}$$

In summary, the DFT is proportional to the set of coefficients of projection onto the sinusoidal basis set, and the inverse DFT is the reconstruction of the original signal as a superposition of its sinusoidal projections. This basic ``architecture'' extends to all linear orthogonal transforms, including wavelets, Fourier transforms, Fourier series, the discrete-time Fourier transform (DTFT), and certain short-time Fourier transforms (STFT).

See §7.1.1 for notation and terminology associated with the DFT.