Fig. 1

Principle of operation of the AAC algorithm in four steps. Top row: content \( {s}_m \) of the segment as a black line on gray background and signal \( {x}_m \) in gray. Bottom row: correlation signal \( {z}_k \) in black and peak detector function \( {y}_k \) in gray. Solid lines show parts of the functions necessary for fundamental frequency estimation in the respective figure. Dotted lines indicate future values and values that have already led to an estimate. Column A: segment \( {s}_m \) is chosen as a slice of the signal \( {x}_m \) with length \( {M}_s \) starting at the first sample. Column B: the shift between both signals is increased, and the correlation signal \( {z}_k \) is calculated following Eq. (1). Column C: the shift between \( {s}_m \) and \( {x}_m \) is further increased to the point where both signals are qualitatively similar, and the signal \( {z}_k \) exhibits a maximum. The displacement of the segment to the signal provides a measure for the period. Column D: after having detected a maximum in \( {z}_k \), the content of the segment is replaced with a slice of the signal \( {x}_m \) starting right after the detected peak, and fundamental frequency detection restarts