Relaxed code-excited linear prediction (RCELP) is a method used in some advanced speech codecs. The RCELP algorithm does not attempt to match the original signal exactly. Instead, it matches a version of this original signal that conforms to a simplified pitch contour.[1]

References

  1. ^ Bäckström, Tom; Fuchs, Guillaume (2017-01-01). Speech Coding. Signals and Communication Technology. Springer International Publishing. pp. 205–215. doi:10.1007/978-3-319-50204-5_14. ISBN 9783319502021.