Log-Mel Spectrogram
A time-frequency representation formed by applying a Mel-scaled filter bank to a spectrum and taking logarithmic energies, widely used as speech-model input.
Speech Processing Context
A log-Mel spectrogram maps short-time spectral power into perceptually motivated Mel bands and then applies logarithmic compression. It is a common acoustic front end for modern ASR systems, where sample rate, window, hop, FFT size, Mel-band count, and normalization form part of the model input contract.
Representation Boundary
A log-Mel spectrogram and MFCC are not identical. MFCC pipelines normally apply an additional DCT to log-Mel energies and retain a selected number of cepstral coefficients.
Related Speech Processing Concepts
Related technical article: Audio Activity and Anomaly Detection.