iracema.harmonics

This module contains the implementation of a harmonic extraction model for monophonic audio.

iracema.harmonics.extract(fft, pitch, nharm=16, minf0=24, maxf0=4200, perc_tol=0.04)

Extract the harmonics of an audio signal, given its fft and pitch curve, using nharm harmonics. The pitch detection method will search for a a fundamental frequency in the interval between minf0 and maxf0.

Parameters

• fft (FFT) – FFT time series.

• pitch (TimeSeries) – Pitch time series.

• nharm (int) – Number of harmonics to extract.

• minf0 (float) – Minimum fundamental frequency.

• maxf0 (float) – Maximum fundamental frequency.

• min_mag_cand (float) – Minimum amplitude to be candidate, relative to the frame’s peak.

• noisiness_thresh (float) – Upper noisiness limit for each frame. If the noisiness exceeds this value, the algorithm will consider that the frame has no fundamental frequency.

• perc_tol (float) – Frequency tolerance used to search for peaks in the region around the expected frequency of each harmonic.

Returns

harmonics – A dictionary containing the results of the harmonics extraction, composed of three TimeSeries objects, which can be accessed using the following keys: ‘frequency’, ‘magnitude’and ‘phase’.

Return type

dictionary

iracema.harmonics.frame_harmonics(frame_pitch, fft_frame_mag, fft_frame_phase, fft_max_freq, nharm, perc_tol)

Extract the harmonics for a fft frame.