spafe.frequencies.fundamental_frequencies

Credits to:

Patrice Guyot. (2018, April 19). Fast Python implementation of the Yin algorithm (Version v1.1.1). Zenodo. http://doi.org/10.5281/zenodo.1220947

class spafe.frequencies.fundamental_frequencies.FundamentalFrequenciesExtractor(debug=False)

Bases: object

compute_yin(sig, fs, dataFileName=None, w_len=512, w_step=256, f0_min=50, f0_max=3000, harmo_thresh=0.1)

Compute the Yin Algorithm. Return fundamental frequency and harmonic rate.

Parameters:

• sig (list) – Audio signal (list of float)
• fs (int) – sampling rate (= average number of samples pro 1 second)
• w_len (int) – size of the analysis window (in #samples)
• w_step (int) – size of the lag between two consecutives windows (in #samples)
• f0_min (int) – Minimum fundamental frequency that can be detected (in Hertz)
• f0_max (int) – Maximum fundamental frequency that can be detected (in Hertz)
• harmo_tresh (int) – Threshold of detection. The yalgorithmù return the first minimum of the CMND fubction below this threshold.

Returns:

tuple include the following

• pitches : list of fundamental frequencies,

• harmonic_rates: list of harmonic rate values for each fundamental

  frequency value (= confidence value)

• argmins : minimums of the Cumulative Mean Normalized DifferenceFunction
• times : list of time of each estimation

Return type:

(tuple)

cumulativeMeanNormalizedDifferenceFunction(df, N)

Compute cumulative mean normalized difference function (CMND). This corresponds to equation (8) in [1].

Parameters:

• df (list) – Difference function
• N (int) – length of data
• tau_max (int) – integration window size

Returns:

cumulative mean normalized difference function

Return type:

(list)

differenceFunction(x, N, tau_max)

Compute difference function of data x. This corresponds to equation (6) in [1] Fastest implementation. Use the same approach than differenceFunction_scipy. This solution is implemented directly with np fft.

Parameters:

• x (array) – audio data
• N (int) – length of data
• tau_max (int) – integration window size

Returns:

difference function

Return type:

(list)

getPitch(cmdf, tau_min, tau_max, harmo_th=0.1)

Return fundamental period of a frame based on CMND function.

• cmdf: Cumulative Mean Normalized Difference function

Parameters:

• tau_min (int) – minimum period for speech
• tau_max (int) – maximum period for speech
• harmo_th (float) – harmonicity threshold to determine if it is necessary to compute pitch frequency

Returns:

fundamental period if there is values under threshold, 0 otherwise

Return type:

(float)

main(sig, fs, w_len=1024, w_step=256, f0_min=70, f0_max=200, harmo_thresh=0.85, audioDir='./', dataFileName=None)

Run the computation of the Yin algorithm on a example file.

Parameters:

• sig (list) – Audio signal (list of float)
• fs (int) – sampling rate (= average number of samples pro 1 second)
• w_len (int) – size of the analysis window (in #samples)
• w_step (int) – size of the lag between two consecutives windows (in #samples)
• f0_min (int) – Minimum fundamental frequency that can be detected (in Hertz)
• f0_max (int) – Maximum fundamental frequency that can be detected (in Hertz)
• harmo_tresh (int) – Threshold of detection. The yalgorithmù return the first minimum of the CMND fubction below this threshold.

Returns:

tuple include the following

• pitches : list of fundamental frequencies,

• harmonic_rates: list of harmonic rate values for each

  fundamental frequency value (= confidence value)

• argmins : minimums of the Cumulative Mean Normalized DifferenceFunction
• times : list of time of each estimation

Return type:

(tuple)