What Is Colored Noise?

Colored noise is a category of random noise signals characterized by a non-uniform power spectral density, meaning that power is not distributed equally across frequencies. The term takes its name from analogy with optical color: just as visible colors correspond to different wavelength distributions of light, noise "colors" correspond to different frequency distributions of power. The most studied colored noise types are white, pink, and brown (also called red or Brownian) noise, each defined by a specific relationship between power and frequency.

The classification by color provides a concise way to describe the statistical properties of a noise process relevant to its behavior in physical systems, electronic circuits, biological signals, and communication channels. Understanding which noise color is present in a system is a practical prerequisite for designing filters, estimators, and detection algorithms that account for the actual stochastic environment.

White, Pink, and Brown Noise

White noise has a flat power spectral density across all frequencies, making it the reference case. Pink noise, also called 1/f noise, has a power spectral density that decreases by 3 dB per octave as frequency increases, so that equal-width logarithmic frequency bands carry equal power. Brown noise, derived from Brownian motion, has a power spectral density that falls by 6 dB per octave, proportional to 1/f². Blue noise increases at 3 dB per octave and violet noise at 6 dB per octave. An analysis of colored noise in computational inference for neurophysiological time series illustrates how misidentifying the noise color in fMRI data can lead to incorrect statistical conclusions when hypothesis testing assumes white noise residuals. The correct characterization of noise color is therefore a prerequisite for valid inference.

Generation and Modeling

Colored noise is generated analytically by shaping white noise through a filter whose frequency response imposes the desired spectral slope. A pink noise generator, for instance, applies a filter with a -10 dB per decade roll-off to a white noise source. For discrete-time systems, autoregressive and moving-average (ARMA) models represent colored noise as the output of a rational transfer function driven by white noise. This representation is foundational to the Kalman filter and related optimal estimators, which require a complete noise model including spectral color. An IEEE Xplore paper on adaptive detection for unknown noise power spectral densities addresses the problem of detecting signals embedded in colored noise whose spectral shape is not fully known in advance.

Role in Physical and Biological Systems

Pink (1/f) noise appears in an unusually wide range of physical, biological, and economic phenomena, from electronic device current fluctuations to human heartbeat interval variability and neuroscience measurements. This ubiquity has generated substantial theoretical interest in identifying a universal mechanism. An arXiv preprint examining the color of neural noise reviews evidence for 1/f-type spectra in neural recordings and discusses competing generative models. Brown noise arises naturally in random-walk processes and is observed in phenomena such as stock price changes, diffusion signals, and certain seismic records. The color of the noise background affects the threshold for detecting weak signals, a consideration in radar, sonar, and gravitational-wave detector design.

Applications

Colored noise has applications in a wide range of fields, including:

• Electronic circuit design, where 1/f noise sets the fundamental sensitivity floor for low-frequency amplifiers and oscillators
• Communications systems, where channel noise color shapes equalizer and detector design
• Biomedical signal processing, where neural and cardiac signals exhibit colored noise backgrounds that affect detection of clinical events
• Audio engineering and acoustics, where pink noise serves as a standard stimulus for speaker and room testing
• Geophysics and seismology, where noise color characterization aids in separating signal from background