Real-time pitch correction in C++

Abstract

This paper presents an implementation approach for real-time pitch correction in mon-ophonic digital audio using C++. The described approach is based on the original, now expired Auto-Tune patent filed in 1998 by Harold A. Hildebrand and its main applica-tion is intonation correction in vocal performances.

The presented pitch correction method uses autocorrelation and mean squared error to find the input signal’s fundamental frequency. After a successful pitch detection, correction to a desired musical note is performed by resampling the signal with a cir-cular buffer.

The pitch correction algorithm was implemented to a market-ready audio plugin for the client of this thesis work—Sixth Sample Ky—using C++ audio application framework JUCE. The finished product was also evaluated against Auto-Tune.

The resulting audio plugin is a competitive, although simplified alternative to industry-leading pitch correction software. Minor improvement possibilities remain in terms of resistance to noise and pitch correction composure with certain types of source audio.