Mel-Frequency Cepstral Coefficients
Mel-Frequency Cepstral Coefficients — Classical speech features derived from log Mel-filterbank energies followed by a cepstral transform.
Technical Context
Classical speech features derived from log Mel-filterbank energies followed by a cepstral transform.
MFCC compresses phase and detailed spectral structure; windowing, filter-bank design, coefficient count, and normalisation affect downstream performance.
Related Articles
Relationship to the MFCC Abbreviation
Mel-Frequency Cepstral Coefficients is commonly abbreviated as MFCC; the two names denote the same feature family rather than different algorithms. The expanded form makes the Mel scale and cepstral coefficients explicit, while the acronym is more common in day-to-day technical use.
Applying a logarithm to Mel-filterbank energies and then using a DCT produces a compact cepstral representation. Coefficient count, liftering, cepstral mean/variance normalization, and delta features are pipeline choices rather than properties implied by the acronym itself.
Related: Cepstrum, Mel Scale, Filter Bank, Speech Feature Extraction with MFCC.