marytts.htsengine

Class HTSVocoder

java.lang.Object

marytts.htsengine.HTSVocoder

public class HTSVocoder
extends Object

Synthesis of speech out of speech parameters. Mixed excitation MLSA vocoder. Java port and extension of HTS engine API version 1.04 Extension: mixed excitation

Author:

Marcela Charfuelan

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	HTSVocoder.HTSVocoderDataProducer

Field Summary

Fields

Modifier and Type	Field and Description
static int	IPERIOD
static int	IRLENG
static double	LZERO
static int	PADEORDER
static int	SEED
static double	ZERO

Constructor Summary

Constructors

Constructor and Description
HTSVocoder()

Method Summary

Methods

Modifier and Type	Method and Description
static double	b2en(double[] b, int m, double a) b2en: functions for postfiltering
static void	b2mc(double[] b, double[] mc, int m, double a) b2mc: transform MLSA digital filter coefficients to mel-cepstrum
static void	c2ir(double[] c, int nc, double[] hh, int leng) c2ir: The minimum phase impulse response is evaluated from the minimum phase cepstrum
static void	freqt(double[] c1, int m1, double[] c2, int m2, double a) freqt: frequency transformation
static void	gc2gc(double[] c1, int m1, double g1, double[] c2, int m2, double g2) gc2gc: generalized cepstral transformation
static double[]	genPulseFromFourierMag(double[] mag, double f0) Generate one pitch period from Fourier magnitudes
static double[]	genPulseFromFourierMag(HTSPStream mag, int n, double f0)
static AudioFormat	getHTSAudioFormat(HMMData htsData) get the audio format produced by the hts vocoder
static void	gnorm(double[] c1, double[] c2, int m, double g) ignorm: gain normalization
AudioInputStream	htsMLSAVocoder(HTSParameterGeneration pdf2par, HMMData htsData) HTS_MLSA_Vocoder: Synthesis of speech out of mel-cepstral coefficients.
double[]	htsMLSAVocoder(HTSPStream lf0Pst, HTSPStream mcepPst, HTSPStream strPst, HTSPStream magPst, boolean[] voiced, HMMData htsData, HTSVocoder.HTSVocoderDataProducer audioProducer)
static void	htsMLSAVocoderCommand(String[] args) Stand alone vocoder reading parameters from files in SPTK format, parameters in args[] array in the following order:
static void	ignorm(double[] c1, double[] c2, int m, double ng) ignorm: inverse gain normalization
static void	lsp2lpc(double[] lsp, double[] a, int m) lsp2lpc: transform LSP to LPC.
void	lsp2mgc(double[] lsp, double[] mgc, int m, double alpha) lsp2mgc: transform LSP to MGC.
static void	main(String[] args)
static void	main1(String[] args) Stand alone testing reading parameters from files in SPTK format
static void	mc2b(double[] mc, double[] b, int m, double a) mc2b: transform mel-cepstrum to MLSA digital filter coefficients
static void	mgc2mgc(double[] c1, int m1, double a1, double g1, double[] c2, int m2, double a2, double g2) mgc2mgc: frequency and generalized cepstral transformation
static double	mglsadf(double x, double[] b, int m, double a, int n, double[] d) mglsadff: sub functions for MGLSA filter
static double	mlsadf(double x, double[] b, int m, double a, double[] d, int pt2, int[] pt3) mlsadf: HTS Mel Log Spectrum Approximation filter
static void	postfilter_mgc(double[] mgc, int m, double alpha, double beta) posfilter: postfilter for mel-cepstrum.
double	uniformRand() uniform_rand: generate uniformly distributed random numbers 1 or -1
static void	vocoderList(String[] args)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

IPERIOD

public static final int IPERIOD

See Also:

Constant Field Values

SEED

public static final int SEED

See Also:

Constant Field Values

PADEORDER

public static final int PADEORDER

See Also:

Constant Field Values

IRLENG

public static final int IRLENG

See Also:

Constant Field Values

ZERO

public static final double ZERO

See Also:

Constant Field Values

LZERO

public static final double LZERO

See Also:

Constant Field Values

Constructor Detail

HTSVocoder

public HTSVocoder()

Method Detail

htsMLSAVocoder

public AudioInputStream htsMLSAVocoder(HTSParameterGeneration pdf2par,
                              HMMData htsData)
                                throws Exception

HTS_MLSA_Vocoder: Synthesis of speech out of mel-cepstral coefficients. This procedure uses the parameters generated in pdf2par stored in: PStream mceppst: Mel-cepstral coefficients PStream strpst : Filter bank stregths for mixed excitation PStream magpst : Fourier magnitudes PStream lf0pst : Log F0

Parameters:

pdf2par - pdf2par

htsData - htsData

Returns:

DDSAudioInputStream

Throws:

Exception - Exception

getHTSAudioFormat

public static AudioFormat getHTSAudioFormat(HMMData htsData)

get the audio format produced by the hts vocoder

Parameters:

htsData - htsData

Returns:

audioformat

htsMLSAVocoder

public double[] htsMLSAVocoder(HTSPStream lf0Pst,
                      HTSPStream mcepPst,
                      HTSPStream strPst,
                      HTSPStream magPst,
                      boolean[] voiced,
                      HMMData htsData,
                      HTSVocoder.HTSVocoderDataProducer audioProducer)
                        throws Exception

Throws:

Exception

mlsadf

public static double mlsadf(double x,
            double[] b,
            int m,
            double a,
            double[] d,
            int pt2,
            int[] pt3)

mlsadf: HTS Mel Log Spectrum Approximation filter

Parameters:

x - x

b - b

m - m

a - a

d - d

pt2 - pt2

pt3 - pt3

Returns:

x

uniformRand

public double uniformRand()

uniform_rand: generate uniformly distributed random numbers 1 or -1

Returns:

rand.nextboolean

mc2b

public static void mc2b(double[] mc,
        double[] b,
        int m,
        double a)

mc2b: transform mel-cepstrum to MLSA digital filter coefficients

Parameters:

mc - mc

b - b

m - m

a - a

b2mc

public static void b2mc(double[] b,
        double[] mc,
        int m,
        double a)

b2mc: transform MLSA digital filter coefficients to mel-cepstrum

Parameters:

b - b

mc - mc

m - m

a - a

freqt

public static void freqt(double[] c1,
         int m1,
         double[] c2,
         int m2,
         double a)

freqt: frequency transformation

Parameters:

c1 - c1

m1 - m1

c2 - c2

m2 - m2

a - a

c2ir

public static void c2ir(double[] c,
        int nc,
        double[] hh,
        int leng)

c2ir: The minimum phase impulse response is evaluated from the minimum phase cepstrum

Parameters:

c - c

nc - nc

hh - hh

leng - leng

b2en

public static double b2en(double[] b,
          int m,
          double a)

b2en: functions for postfiltering

Parameters:

b - b

m - m

a - a

Returns:

en

ignorm

public static void ignorm(double[] c1,
          double[] c2,
          int m,
          double ng)

ignorm: inverse gain normalization

Parameters:

c1 - c1

c2 - c2

m - m

ng - ng

gnorm

public static void gnorm(double[] c1,
         double[] c2,
         int m,
         double g)

ignorm: gain normalization

Parameters:

c1 - c1

c2 - c2

m - m

g - g

lsp2lpc

public static void lsp2lpc(double[] lsp,
           double[] a,
           int m)

lsp2lpc: transform LSP to LPC. lsp[1..m] → a=lpc[0..m] a[0]=1.0

Parameters:

lsp - lsp

a - a

m - m

gc2gc

public static void gc2gc(double[] c1,
         int m1,
         double g1,
         double[] c2,
         int m2,
         double g2)

gc2gc: generalized cepstral transformation

Parameters:

c1 - c1

m1 - m1

g1 - g1

c2 - c2

m2 - m2

g2 - g2

mgc2mgc

public static void mgc2mgc(double[] c1,
           int m1,
           double a1,
           double g1,
           double[] c2,
           int m2,
           double a2,
           double g2)

mgc2mgc: frequency and generalized cepstral transformation

Parameters:

c1 - c1

m1 - m1

a1 - a1

g1 - g1

c2 - c2

m2 - m2

a2 - a2

g2 - g2

lsp2mgc

public void lsp2mgc(double[] lsp,
           double[] mgc,
           int m,
           double alpha)

lsp2mgc: transform LSP to MGC. lsp=C[0..m] mgc=C[0..m]

Parameters:

lsp - lsp

mgc - mgc

m - m

alpha - alpha

mglsadf

public static double mglsadf(double x,
             double[] b,
             int m,
             double a,
             int n,
             double[] d)

mglsadff: sub functions for MGLSA filter

Parameters:

x - x

b - b

m - m

a - a

n - n

d - d

Returns:

x

postfilter_mgc

public static void postfilter_mgc(double[] mgc,
                  int m,
                  double alpha,
                  double beta)

posfilter: postfilter for mel-cepstrum. It uses alpha and beta defined in HMMData

Parameters:

mgc - mgc

m - m

alpha - alpha

beta - beta

genPulseFromFourierMag

public static double[] genPulseFromFourierMag(HTSPStream mag,
                              int n,
                              double f0)

genPulseFromFourierMag

public static double[] genPulseFromFourierMag(double[] mag,
                              double f0)

Generate one pitch period from Fourier magnitudes

Parameters:

mag - mag

f0 - f0

Returns:

pulse

main1

public static void main1(String[] args)
                  throws IOException,
                         InterruptedException,
                         Exception

Stand alone testing reading parameters from files in SPTK format

Parameters:

args - args

Throws:

IOException - IOException

InterruptedException - InterruptedException

Exception - Exception

htsMLSAVocoderCommand

public static void htsMLSAVocoderCommand(String[] args)
                                  throws IOException,
                                         InterruptedException,
                                         Exception

Stand alone vocoder reading parameters from files in SPTK format, parameters in args[] array in the following order:

The type of spectrum parameters is set through the parameters gamma and alpha

Parameters:

args -

example iput parameters:

0 0.45 0 0.0 16000 80 cmu_us_arctic_slt_a0001.mgc 75 cmu_us_arctic_slt_a0001.lf0 3 vocoder_out.wav cmu_us_arctic_slt_a0001.str 15 mix_excitation_filters.txt 5 48 cmu_us_arctic_slt_a0001.mag 30

example input parameters without mixed excitation:

0 0.45 0 0.0 16000 80 cmu_us_arctic_slt_a0001.mgc 75 cmu_us_arctic_slt_a0001.lf0 3 vocoder_out.wav

Throws:

IOException - IOException

InterruptedException - InterruptedException

Exception - Exception

main

public static void main(String[] args)
                 throws IOException,
                        InterruptedException,
                        Exception

Throws:

IOException

InterruptedException

Exception

vocoderList

public static void vocoderList(String[] args)
                        throws IOException,
                               InterruptedException,
                               Exception

Throws:

IOException

InterruptedException

Exception