Glossary
Every defined term in this course, A–Z.
Absolutely integrable ($L^1(\mathbb R)$)
Read the section →\int_{-\infty}^\infty|f(t)|\,dt is finite -- a stronger, whole-real-line condition than being L^1 on a bounded interval.
Aliasing
Read the section →Two different frequencies producing the identical sample sequence at a given sampling rate.
Angular frequency ($\omega$)
Read the section →The variable \omega a Fourier transform is a function of, in radians per unit time.
Anti-aliasing filter
Read the section →A filter applied before sampling to force a real signal's spectrum closer to band-limited.
Apparent frequency
Read the section →The one frequency in [0,f_s/2] that a reconstruction from aliased samples will actually produce.
Band-limited (to $B$ Hz)
Read the section →A signal whose spectrum is exactly zero past some finite frequency B.
Best approximation property
Read the section →The Fourier partial sum is the closest possible trigonometric polynomial of its degree, in the L^2 sense.
Bit-reversal (order)
Read the section →The fixed input reordering, a byproduct of splitting by successive bits, that the FFT's recursion needs.
Boundary value problem
Read the section →A differential equation with conditions pinned at two points of the domain, not one starting point.
Butterfly (operation)
Read the section →The two-input, one-multiply, two-output primitive that is the FFT's entire arithmetic content.
Cesàro (Fejér) mean
Read the section →The running average of a Fourier series' first several partial sums, used to kill the Gibbs ripple.
Circulant matrix
Read the section →A matrix built from one sequence, where every row is the row above it, shifted and wrapped around.
Circular convolution ($c\circledast x$)
Read the section →The wraparound version of convolution for two finite sequences, computed by a circulant matrix.
Complete (orthogonal system)
Read the section →An orthogonal system with no nonzero vector orthogonal to every member of it.
Convolution
Read the section →A way to combine two functions, (f*g)(t)=\int f(\tau)g(t-\tau)\,d\tau, that is neither their sum nor their product.
DC component
Read the section →The zero-frequency (\omega=0) spike in a signal's spectrum, corresponding to its constant part.
Decomposition
Read the section →Writing a complicated periodic function as a sum of a few simple periodic pieces.
Dense (subset)
Read the section →A subset of a normed space where every element of the space can be approximated arbitrarily well.
Dirac delta ($\delta(t)$)
Read the section →A generalized function, the limit of ever-narrower unit-area pulses, defined by what it does inside an integral.
Dirichlet's conditions
Read the section →The three conditions on a periodic function that guarantee its Fourier series converges.
Energy (of a function)
Read the section →The informal name signal processing gives to \|f\|_2^2.
Fourier cosine series
Read the section →The even-function mirror image of the Fourier sine series -- pure cosines plus a constant.
Fourier sine series
Read the section →The pure-sine Fourier series an odd periodic function's expansion reduces to.
Fourier transform
Read the section →The frequency-domain function \hat f(\omega)=\int_{-\infty}^\infty f(t)e^{-i\omega t}\,dt built from a non-periodic f.
Frequency bin (bin index)
Read the section →One output value X_k of a DFT, indexed by a slot number k, not a frequency in hertz on its own.
Frequency mask
Read the section →A function multiplied against a shifted 2D spectrum to keep or zero out frequency content by region.
Frequency response ($H(\omega)$)
Read the section →The function H(\\omega) that specifies a filter, multiplying an input's transform to produce the output's.
Frequency spectrum (continuous case)
Read the section →The graph of |\hat f(\omega)| against \omega, for a non-periodic signal.
Frequency vector ($\omega_x,\omega_y$)
Read the section →The pair of frequencies in a 2D transform, describing both how fast and which way a pattern oscillates.
Frequency-domain
Read the section →A plot of how much of each frequency is mixed into a signal.
Gibbs constant
Read the section →The universal overshoot fraction, about 8.95% of the jump size, in the Gibbs phenomenon.
Gibbs phenomenon
Read the section →A Fourier partial sum's persistent overshoot near a jump, which never shrinks away as more terms are added.
Half-range expansion
Read the section →The sine or cosine series obtained by extending a function, defined only on half an interval, oddly or evenly.
Harmonic number
Read the section →How many full cycles a term completes while the fundamental completes one, in A\sin(nx+\varphi).
Hilbert space
Read the section →A complete inner product space -- every Cauchy sequence in it converges to a limit still inside it.
Hyperbolic functions ($\cosh$, $\sinh$)
Read the section →\cosh z=\tfrac12(e^z+e^{-z}) and \sinh z=\tfrac12(e^z-e^{-z}), a convenient exponential basis.
Impulse response ($h=\mathcal F^{-1}[H]$)
Read the section →The time-domain function whose Fourier transform is a filter's frequency response.
Impulse train
Read the section →An infinite row of evenly spaced unit Dirac impulses, used to model sampling as multiplication.
Inner product space
Read the section →A vector space equipped with an inner product, generalizing the dot product.
Inverse Fourier transform
Read the section →The formula that recovers f back from its Fourier transform \hat f.
Magnitude spectrum and phase spectrum
Read the section →The size |\hat f(\omega)| and angle \arg\hat f(\omega) of a complex-valued spectrum, plotted separately.
Normal mode
Read the section →A single separated wave-equation solution -- a standing wave with a fixed shape that only grows or shrinks in place.
Nyquist frequency
Read the section →A property of a chosen sampling rate: half of f_s, the highest frequency that rate can represent without aliasing.
Nyquist rate
Read the section →A property of a signal: the minimum sampling rate, 2B, the theorem requires for its bandwidth B.
Overtone
Read the section →A harmonic above the fundamental, numbered starting at 1 -- so overtone number always equals harmonic number minus one.
Parallelogram law
Read the section →The identity \|f+g\|^2+\|f-g\|^2=2\|f\|^2+2\|g\|^2, true only for norms that come from an inner product.
Periodic extension
Read the section →A function's shape copied every T to build a periodic signal out of a non-periodic slice.
Periodization (of a spectrum)
Read the section →The tiling effect sampling in time has on a signal's spectrum in frequency.
Piecewise smooth (function)
Read the section →A function meeting all three of Dirichlet's conditions on one period.
Pointwise convergence
Read the section →A sequence of functions converges at every point separately, with no shared rate.
Radix-2 (FFT)
Read the section →The version of the Cooley-Tukey FFT that splits by 2 at every level, requiring N a power of 2.
Root of unity
Read the section →One of N equally spaced points on the unit circle, the exponential kernel a DFT sum is built from.
Sampling (and sampling interval)
Read the section →Recording a continuous signal's value only at evenly spaced instants, \Delta t apart.
Separable (function)
Read the section →A function of the form g(x)h(y), whose 2D transform is the product of the two functions' 1D transforms.
Sifting property
Read the section →Integrating \delta(t-a) against f returns f(a) -- the defining rule for the Dirac delta.
Sinc function
Read the section →The \sin-over-argument ripple shape, \sin(a\omega)/\omega, that is a rectangular pulse's Fourier transform.
Spatial frequency
Read the section →How fast intensity changes per unit distance across an image -- independent of how bright the image is.
Standing wave (node and antinode)
Read the section →A wave whose spatial shape never travels -- it only grows, shrinks, and reverses in place.
Sturm–Liouville problem
Read the section →The general pattern of eigenvalues and orthogonal eigenfunctions this chapter keeps rediscovering by hand.
Thermal diffusivity ($\alpha$)
Read the section →The material constant governing how fast heat spreads through a rod.
Timbre
Read the section →What makes the same musical pitch sound different on different instruments -- the relative mix of harmonic amplitudes.
Time-domain
Read the section →The ordinary graph of a signal's value against time.
Trigonometric polynomial
Read the section →A finite sum a_0+\sum_{n=1}^N(a_n\cos(n\omega_0x)+b_n\sin(n\omega_0x)).
Twiddle factor
Read the section →The factor W_N^k=e^{-i2\pi k/N} that scales the odd half's DFT before combining it with the even half's.
Uniform convergence
Read the section →A sequence of functions converges to its limit at the same rate everywhere on the domain.