Skip to main content
Fig. 4 | EURASIP Journal on Audio, Speech, and Music Processing

Fig. 4

From: From raw audio to a seamless mix: creating an automated DJ system for Drum and Bass

Fig. 4

Overview of the downbeat tracking algorithm. a The audio waveform, annotated with beats and downbeats. b Isolated features are extracted from the beats. c Isolated features are combined to create contextual features. d The machine learning model classifies each beat, estimating the log-probability of a beat being either the 1st, 2nd, 3rd, or 4th beat in a measure. e The downbeats are determined by summing the log-probabilities along each possible trajectory and choosing the most likely one. Here, the second trajectory is highlighted

Back to article page