NeuroDSP Glossary

The following is a glossary of neuroscience and digital processing related terms that are used in NeuroDSP, as well as links to some external resources for learning DSP.



Properties or components of a signal that are rhythmic.


Properties or components of a signal that are arrhythmic, with no characteristic frequency.

Digital Signal Processing

Digital Signal Processing (DSP) is an area of science & engineering related to the computational analysis of digital signals.

For a collection of openly available resources for learning DSP, check out this list.

If you want to start with a general introduction to DSP, check out Seeing Circles Sines and Signal or for more in-depth descriptions, check out The Scientist and Engineers Guide to Digital Signal Processing.

time domain

Signals that are represented as variations over time, and analyses of such signals.

frequency domain

Signals that are represented in terms of frequencies, and analyses of such signals.

sampling rate

The rate at which samples are taken.

temporal resolution

The precision of a measurement, in the time domain. This is set by the magnitude of time between successive measurements (e.g. 0.01 seconds between samples).

frequency resolution

The precision of a measurement, in the frequency domain. This is set by the magnitude of frequency between successive measurements (e.g. 0.5 Hz between measurements).


Hertz (Hz)

A unit of frequency, as the number of cycles per second.

Decibels (dB)

A unit of intensity, on a logarithmic scale.

Volts (V)

A unit of voltage, typically in the microvolt (uV) range for neural time series.


There are some available (pay-walled) articles that present overviews and guides to filters, including:

  • this guide on using filters for electrophysiological data

  • this primer on when, how, and why to use filters

For an open, in depth, and code-driven tutorial, check out the MNE Filtering Tutorial.

Impulse Response

The response of a filter when presented with an impulse; a single, brief input.


A Finite Impulse Response filter, meaning its impulse response settles to zero in finite time.


An Infinite Impulse Response filter, meaning the filter is recursive, and its impulse response continues infinitely.


The range (band) of frequencies that are unattenuated by a filter.


The range (band) of frequencies that are attenuated (stopped) by a filter.


The type of filter, defined in terms of what frequency bands or ranges it passes through, or filters out.

  • bandpass: a filter whose passband is a specific frequency band, bound by a low and high frequency point.

  • bandstop: a filter that passes through all frequencies except a band region that is attenuated.

  • lowpass: a filter whose passband is all frequencies below a filter frequency (low frequencies pass through).

  • highpass: a filter whose passband is all frequencies above a filter frequency (high frequencies pass through).

transition band

The range of frequencies that are in the transition region between the passband and the stopband.

frequency response

The response profile of a filter, specifying the gain and phase shift applied by the filter at each frequency.

Rhythms & Bursts


Periodic activity that lasts for a short or transient amount of time, as in a ‘burst of oscillatory activity’.

Time Frequency

We currently have two general approaches to time frequency analyses:


The number of occurrences over a unit of time, typically referred to as cycles per second, and measured in Hz.


The position, at a point in time, on a waveform cycle.


The magnitude of a signal, as the peak-to-trough distance.


The squared magnitude of a signal.


A single cycle of a rhythm, defined as the time between two consecutive troughs (or peaks).

hilbert transform

A mathematical transform that computes the ‘analytic signal’, a complex-valued representation of a time-series (signal) that can be used to find its analytic amplitude and phase.


A wave-like signal, or ‘brief oscillation’, that starts at zero amplitude, increases in amplitude to some value, and then decays back to zero.


Many of the spectral methods available are based on the Fourier transform, for which there is an interactive guide by Better Explained and an explainer video by 3Blue1Brown.

Fourier transform

A mathematical transformation to decompose a time series into a frequency representation.

power spectrum

A frequency domain representation, as an estimate of the power across frequencies in a signal.

median filter

A smoothing approach to replace each value in a signal with the median of the neighboring entries.

coefficient of variation

A standardized measure of dispersion, as the ratio of the standard deviation to the mean.


For an overview of the aperiodic signals available in terms of their 1/f characteristics, check out this article from scholarpedia.

noise signal

Formally, a noise signal is a signal produced by a stochastic (random) process. The aperiodic signals that are simulated in NeuroDSP are, technically, noise signals.


A relationship between two quantities, whereby one quantity varies as a power of another. One-over-f relationships are powerlaw, as the spectral power varies by a power of the frequency.

1/f signal

A signal for which the power spectrum can be described by a powerlaw of the form \(1/f^\chi\), where \(\chi\) refers to the exponent of the powerlaw.

colored noise

The ‘color’ of noise refers to the 1/f exponent of the power spectrum of a noise signal.

  • white noise: a signal with a \(1/f^0\) power spectrum, which is flat with equal power across all frequencies

  • pink noise: a signal with a \(1/f^1\) power spectrum

  • brown noise: a signal with a \(1/f^2\) power spectrum, sometimes also known as red noise

random walk

A random process that describes a path of a succession of random steps.