A mel-frequency scale is a perceptual frequency scale used in audio signal processing and psychoacoustics that approximates human pitch perception. It maps physical frequency in hertz to mel units using the transformation: mel = 2595 × log₁₀(1 + f/700), where f is the frequency in Hz. Below 1000 Hz the scale is approximately linear with a slope of ~1 mel/Hz; above 1000 Hz it becomes logarithmic, reflecting the ear's diminishing frequency resolution. The scale is defined by three parameters: (1) the reference formula mapping Hz to mels, (2) the inflection point at 1000 Hz where linearity gives way to logarithmic compression, and (3) the normalization constants (2595 and 700) derived from empirical psychoacoustic experiments by Stevens, Volkmann, and Newman (1937) and later refined by Steinberg and Palmer (1993). It persists through standardization in audio toolkits (Librosa, HTK, Praat), speech recognition frameworks, and music information retrieval systems as a reusable perceptual mapping. [formal: mel | substrate: mind | horizon: hours | explicit: yes | epoch: 0.01]
Accepted ontology entry
mel-frequency-scale
A mel-frequency scale is a perceptual frequency scale used in audio signal processing and psychoacoustics that approximates human pitch perception. It maps physical frequency in hertz to mel units using the transformation: mel = 2595 × log…
Definition
Why it is in scope
A human-made perceptual frequency scale that maps physical sound frequency (Hz) onto a scale linear below 1000 Hz and logarithmic above, approximating human pitch perception. It is a mathematical transformation prescribed by psychoacoustic research and embedded in audio processing algorithms, speech recognition pipelines, and music information retrieval systems — a standardized conceptual tool built to persist across disciplines.
Names and aliases
- mel-frequency-scaleen · CANONICAL
Relations from this entry
- cmspywurr06ztjlssi76045rfSERVES →
The mel-frequency scale was designed to approximate human auditory perception, making it the standard frequency representation for speech recognition feature extraction (MFCCs, filter banks). It serves speech-recognition's need for perception-matched frequency analysis.
- cmsqz0txz004nswyn3j0xokafINSTANCE_OF →
Mel-frequency scale is a specific kind of frequency scale: it maps physical frequency to a perceptual scale using the mel formula. A competent speaker would call mel-frequency scale a frequency scale. Specific to general — correct for Law 9.
Relations to this entry
- cmspw3d3806pzjlss3rcmuezg← DERIVED_FROM
Which-came-first: the mel-frequency-scale (1930s psychoacoustic research) predates MFCCs (1980s speech recognition). The mel scale is a foundational building block used in computing MFCCs — MFCCs would not exist without the mel-frequency mapping that defines their frequency axis.
- cmsqals7w0004ox1y79lqk3dy← DERIVED_FROM
Which-came-first: the mel-frequency-scale (1930s) predates cepstral-envelope (1990s speaker recognition). Cepstral-envelope computation uses the mel-frequency scale to map spectral content into perceptual bins before extracting the envelope. The mel scale is a foundational step in computing cepstral-envelopes.
- cmsqb8db6002fox1yi8ldf3fg← DERIVED_FROM
which-came-first: mel-frequency-scale (1930s psychoacoustic research) predates mel-frequency-spectrogram (developed for audio/speech processing). A mel-frequency-spectrogram is computed by applying a STFT then warping the frequency axis using the mel-frequency-scale. The mel scale is a foundational building block — the spectrogram would not exist without it.
- cmsr1p0gz0020kp539o6mpun7← DEPENDS_ON
MFCCs needs mel-frequency-scale to operate: the mel filter bank is the first processing stage that converts the linear spectrum to mel-frequency. Remove mel-frequency-scale and MFCCs stop operating. Law 8 removal test.
- cmsqwdhy2004jax3hdwf3yowy← DEPENDS_ON
Mel-spectrogram operates by projecting spectral energy onto mel-frequency bins. Remove mel-frequency-scale and the projection target vanishes — a mel spectrogram cannot be computed without it. The removal test passes: X stops operating without Y.
- cmspxgu9f06vtjlssp8aaz07z← DEPENDS_ON
The mel-frequency filterbank is defined entirely by the mel frequency scale — its bands are spaced according to the mel scale. Remove the mel scale and the filterbank has no purpose or definition. Constitutive present-tense dependency.
- cmspw3d3806pzjlss3rcmuezg← DEPENDS_ON
MFCC computation requires a mel-frequency filterbank to map linear frequency to perceptually scaled bands before cepstral analysis. Remove the mel-frequency scale and the MFCC pipeline loses its perceptual warping and no longer operates as designed.
- cmspxgu9f06vtjlssp8aaz07z← DERIVED_FROM
A mel filterbank implements a set of bandpass filters spaced according to the mel-frequency scale; the filterbank design is historically and conceptually downstream of the scale definition.
Record identity
- Created
- Aug 12, 2026, 5:10 PM UTC
- Content hash
- 9cc92cd8fee83874221631b36888558d37b0af8432ed2836624025359f7f3e75