How Computers Hear Rhythm: Tempo, Beats, and Musical Style

Episoden

1. 1

   Sound, Frames, and the Pulse of Music

   Harmony asks which notes are sounding. Rhythm asks how fast the music moves, and where each beat lands. Discover why a computer prepares audio the same way for both questions—and why timing still needs its own way of listening.

2. 2

   The Onset Envelope – Hearing Attacks Instead of Notes

   Beats hide in the moments when sound gets suddenly stronger. Meet spectral flux: a one-number-per-frame envelope that peaks at note attacks and drum hits—and ignores the endless tiny peaks of the raw wave.

3. 3

   Tempo from Repetition – Autocorrelation and the Octave Trap

   Once you have a map of attacks, tempo is the rate at which that map repeats. Autocorrelation finds the lag that best matches the pulse—while a gentle preference for mid-range speeds, and two different accuracy rules, keep the famous half-time / double-time ambiguity honest.

4. 4

   Placing Every Beat – Regularity You Can Hear

   Tempo is a speed limit. Beat tracking is driving on the road. Dynamic programming lands on strong onsets while staying evenly spaced—and a mixed click track lets your ear, not only a chart, decide whether the groove is right.

5. 5

   Style from Sound – Features, Honest Splits, and Genre

   Tempo is one number. Style is a texture. Learn how 47 compact measurements describe timbre and rhythm, why tempo must be measured on the whole clip, and why every window of a file must stay on one side of the train/test split.