librosax.autocorrelate

autocorrelate(y: Array, *, max_size: int | None = None, axis: int = -1) → Array[source]

Bounded-lag auto-correlation.

This function computes the autocorrelation of a signal using FFT for efficiency. The autocorrelation is bounded to a maximum lag specified by max_size.

Parameters:

  • y – Array to autocorrelate.

  • max_size – Maximum correlation lag. If None, defaults to y.shape[axis] (unbounded).

  • axis – Axis along which to autocorrelate. Default is -1 (last axis).

Returns:

Autocorrelated array. If max_size is specified, the shape along axis will be max_size. Otherwise, it matches y.shape[axis].

Examples

Compute full autocorrelation of a white noise signal

>>> import numpy as np
>>> y = np.random.randn(256)
>>> z = librosax.autocorrelate(y)
>>> z.shape
(256,)

Compute autocorrelation with a maximum lag of 32 samples

>>> z = librosax.autocorrelate(y, max_size=32)
>>> z.shape
(32,)

Autocorrelate along the time axis of a batch of signals

>>> y = np.random.randn(10, 256)  # 10 signals of length 256
>>> z = librosax.autocorrelate(y, axis=-1)
>>> z.shape
(10, 256)