cLpc

Description

This component computes linear predictive coding (LPC) coefficients from PCM frames. Burg's algorithm and the standard ACF/Durbin based method are implemented for LPC coefficient computation. The output of LPC filter coefficients, reflection coefficients, residual signal, and LP spectrum is supported.

Type hierarchy

cDataProcessor → cVectorProcessor → cLpc

Fields

• method (string) [default: acf]

  This option sets the lpc method to use. Choose between: 'acf' acf (autocorrelation) method with Levinson-Durbin algorithm , 'burg' Burg method (N. Anderson (1978))

• p (numeric) [default: 8.0]

  Predictor order (= number of lpc coefficients)

• saveLPCoeff (numeric) [default: 1.0]

  1 = save LP coefficients to output

• lpGain (numeric) [default: 0.0]

  1 = save lpc gain (error) in output vector

• saveRefCoeff (numeric) [default: 0.0]

  1 = save reflection coefficients to output

• residual (numeric) [default: 0.0]

  1 = compute lpc residual signal and store in output frame

• residualGainScale (numeric) [default: 0.0]

  1 = scale lpc residual signal by lpc gain (divides each frame by gain)

• forwardFilter (numeric) [default: 0.0]

  1 = apply forward instead of inverse filter when computing residual

• lpSpectrum (numeric) [default: 0.0]

  1 = compute lp spectrum using 'lpSpecDeltaF' as frequency resolution or 'lpSpecBins' bins

• forwardLPspec (numeric) [default: 1.0]

  1 = compute forward filter transfer function as LP spectrum, instead of old default behaviour of computing spectrum of inverse filter.

• forwardLPspecFloor (numeric) [default: 1e-13]

  Floor value to add to all spectral bins of inverse LP spectrum when inverting the spectrum (1/x) to avoid division by 0. Default: 10^-13.

• lpSpecDeltaF (numeric) [default: 10.0]

  frequency resolution of lp spectrum (only applicable if 'lpSpectrum=1')

• lpSpecBins (numeric) [default: 100.0]

  number of bins to compute lp spectrum for (overrides lpSpecDeltaF) (only applicable if 'lpSpectrum=1')