F0pitch

F0pitch refers to the process and tools used to estimate the fundamental frequency (F0) of a voiced audio signal, effectively tracking pitch over time. It encompasses a range of algorithms, software libraries, and research methods designed to produce a time series of pitch values from speech or singing recordings. A typical F0pitch system also includes voicing decision to distinguish voiced from unvoiced segments and may apply post-processing to produce a smooth pitch contour.

Common approaches in F0pitch fall into time-domain and frequency-domain categories. Time-domain methods analyze waveform periodicity using autocorrelation or average magnitude difference functions (AMDF), with well-known examples such as the YIN algorithm. Frequency-domain methods rely on spectral representations or cepstrum-based techniques to identify the pitch period. Modern implementations often combine features, apply robustness enhancements, and include post-processing steps to handle rapid pitch changes, vibrato, creaky voice, or octave errors. Outputs typically include the estimated F0 trajectory in hertz and a voicing probability or binary decision for each frame.

Applications of F0pitch span linguistic analysis, where pitch contributes to prosody and tone, as well as singing voice synthesis, voice transformation, and speech processing tasks such as transcription or enhancement. Accurate F0 estimation is important for natural-sounding synthesis and for studies of intonation and musical pitch in audio.

Challenges in F0pitch include sensitivity to noise, polyphony or overlapping voices, rapid pitch movement, and non-harmonic voice qualities. Handling unvoiced regions and preventing octave mistakes remain common concerns. See also fundamental frequency, pitch detection, YIN, cepstrum-based methods, SWIPE, and Praat.