Mel spectrum is the spectral representation of a signal obtained by applying a mel-scale triangular filterbank to its power spectrum. Its parameters are: (1) the input power spectrum P(f) giving energy at each linear frequency bin, (2) the mel filterbank matrix H(m,f) mapping linear frequency bins to mel frequency channels, and (3) the number of mel channels determining frequency resolution. Computation: M(m) = sum_f H(m,f) · P(f), yielding energy in each mel-frequency channel. Persistence: stored as a fixed-length vector of mel-channel energies, computable from any power spectrum via a deterministic filterbank application. [formal: spectrum melodicum | substrate: mind | horizon: a moment | explicit: yes | epoch: 0.01]
Accepted ontology entry
mel-spectrum
Mel spectrum is the spectral representation of a signal obtained by applying a mel-scale triangular filterbank to its power spectrum. Its parameters are: (1) the input power spectrum P(f) giving energy at each linear frequency bin, (2) the…
Definition
Why it is in scope
A human-made spectral representation that maps a signal's power spectrum onto the mel frequency scale using a triangular filterbank. It is computed by humans for audio analysis, speech recognition, and music information retrieval — designed to approximate human auditory perception of frequency.
Names and aliases
- mel-spectrumen · CANONICAL
Relations from this entry
- cmspxgu9f06vtjlssp8aaz07zDERIVED_FROM →
The mel-scale filterbank was developed first as a model of human auditory frequency resolution (Byrne and Greenberg, 1981). The mel-spectrum is computed by applying this pre-existing filterbank to a power spectrum. Which-came-first test: filterbank concept predates mel-spectrum as a named construct.
- cmspywurr06ztjlssi76045rfSERVES →
Mel spectrum is computed and maintained for the purpose of furthering speech recognition. Its mel-frequency channelization approximates human auditory perception, making it an effective front-end for speech feature extraction. By design, the mel-spectrum exists to serve speech recognition systems.
- cmspcwtqk04tkjlsslitlqy58DEPENDS_ON →
The mel-spectrum is computed by applying the mel filterbank to a power spectrum. Remove the power spectrum and there is no input to transform — the mel-spectrum computation stops entirely. Present-tense dependency confirmed.
- cmsq4obya07m0jlssh92wv3v0INSTANCE_OF →
The mel-spectrum is a specific type of spectral representation: it displays signal power distributed across the mel frequency scale (a perceptual frequency scale). A competent speaker would call it a spectral representation. This edges mel-spectrum to its nearest kind.
- cmspxgu9f06vtjlssp8aaz07zDEPENDS_ON →
Mel-spectrum is the power spectrum transformed by the mel-filterbank. Remove the mel-filterbank concept and mel-spectrum has no operational definition — it is nothing without the filterbank. Present-tense DEPENDS_ON per Law 8.
Relations to this entry
- cmspw3d3806pzjlss3rcmuezg← DERIVED_FROM
MFCC is computed from the mel-spectrum via discrete cosine transform of its log values. The mel-spectrum concept and computation predates the named MFCC construct (Davis and Mermelstein, 1980). Which-came-first: mel-spectrum existed first and fed into MFCC creation.
Record identity
- Created
- Aug 12, 2026, 1:21 PM UTC
- Content hash
- 335746712f3fce81ef6e8bd5bd9d25994cef1b20a4ba58f2d79891eb8b288e47