third_party/soloud_speech/klatt.h

   1 #ifndef KLATT_H
   2 #define KLATT_H
   3
   4 #include "resonator.h"
   5
   6 #define CASCADE_PARALLEL      1
   7 #define ALL_PARALLEL          2
   8 #define NPAR                 40
   9
  10 class klatt_frame
  11 {
  12 public:
  13         int mF0FundamentalFreq;          // Voicing fund freq in Hz
  14         int mVoicingAmpdb;               // Amp of voicing in dB,            0 to   70
  15         int mFormant1Freq;               // First formant freq in Hz,        200 to 1300
  16         int mFormant1Bandwidth;          // First formant bw in Hz,          40 to 1000
  17         int mFormant2Freq;               // Second formant freq in Hz,       550 to 3000
  18         int mFormant2Bandwidth;          // Second formant bw in Hz,         40 to 1000
  19         int mFormant3Freq;               // Third formant freq in Hz,        1200 to 4999
  20         int mFormant3Bandwidth;          // Third formant bw in Hz,          40 to 1000
  21         int mFormant4Freq;               // Fourth formant freq in Hz,       1200 to 4999
  22         int mFormant4Bandwidth;          // Fourth formant bw in Hz,         40 to 1000
  23         int mFormant5Freq;               // Fifth formant freq in Hz,        1200 to 4999
  24         int mFormant5Bandwidth;          // Fifth formant bw in Hz,          40 to 1000
  25         int mFormant6Freq;               // Sixth formant freq in Hz,        1200 to 4999
  26         int mFormant6Bandwidth;          // Sixth formant bw in Hz,          40 to 2000
  27         int mNasalZeroFreq;              // Nasal zero freq in Hz,           248 to  528
  28         int mNasalZeroBandwidth;         // Nasal zero bw in Hz,             40 to 1000
  29         int mNasalPoleFreq;              // Nasal pole freq in Hz,           248 to  528
  30         int mNasalPoleBandwidth;         // Nasal pole bw in Hz,             40 to 1000
  31         int mAspirationAmpdb;            // Amp of aspiration in dB,         0 to   70
  32         int mNoSamplesInOpenPeriod;      // # of samples in open period,     10 to   65
  33         int mVoicingBreathiness;         // Breathiness in voicing,          0 to   80
  34         int mVoicingSpectralTiltdb;      // Voicing spectral tilt in dB,     0 to   24
  35         int mFricationAmpdb;             // Amp of frication in dB,          0 to   80
  36         int mSkewnessOfAlternatePeriods; // Skewness of alternate periods,   0 to   40 in sample#/2
  37         int mFormant1Ampdb;              // Amp of par 1st formant in dB,    0 to   80
  38         int mFormant1ParallelBandwidth;  // Par. 1st formant bw in Hz,       40 to 1000
  39         int mFormant2Ampdb;              // Amp of F2 frication in dB,       0 to   80
  40         int mFormant2ParallelBandwidth;  // Par. 2nd formant bw in Hz,       40 to 1000
  41         int mFormant3Ampdb;              // Amp of F3 frication in dB,       0 to   80
  42         int mFormant3ParallelBandwidth;  // Par. 3rd formant bw in Hz,       40 to 1000
  43         int mFormant4Ampdb;              // Amp of F4 frication in dB,       0 to   80
  44         int mFormant4ParallelBandwidth;  // Par. 4th formant bw in Hz,       40 to 1000
  45         int mFormant5Ampdb;              // Amp of F5 frication in dB,       0 to   80
  46         int mFormant5ParallelBandwidth;  // Par. 5th formant bw in Hz,       40 to 1000
  47         int mFormant6Ampdb;              // Amp of F6 (same as r6pa),        0 to   80
  48         int mFormant6ParallelBandwidth;  // Par. 6th formant bw in Hz,       40 to 2000
  49         int mParallelNasalPoleAmpdb;     // Amp of par nasal pole in dB,     0 to   80
  50         int mBypassFricationAmpdb;       // Amp of bypass fric. in dB,       0 to   80
  51         int mPalallelVoicingAmpdb;       // Amp of voicing,  par in dB,      0 to   70
  52         int mOverallGaindb;              // Overall gain, 60 dB is unity,    0 to   60
  53         klatt_frame();
  54 };
  55
  56 class darray;
  57 class Element;
  58
  59 class Slope
  60 {
  61 public:
  62         float mValue;                   /* boundary value */
  63         int mTime;                      /* transition time */
  64         Slope()
  65         {
  66                 mValue = 0;
  67                 mTime = 0;
  68         }
  69 };
  70
  71
  72 enum KLATT_WAVEFORM
  73 {
  74         KW_SAW,
  75         KW_TRIANGLE,
  76         KW_SIN,
  77         KW_SQUARE,
  78         KW_PULSE,
  79         KW_NOISE,
  80         KW_WARBLE
  81 };
  82
  83 class klatt
  84 {
  85         // resonators
  86         resonator mParallelFormant1, mParallelFormant2, mParallelFormant3,
  87                       mParallelFormant4, mParallelFormant5, mParallelFormant6,
  88                       mParallelResoNasalPole, mNasalPole, mNasalZero,
  89                           mCritDampedGlotLowPassFilter, mDownSampLowPassFilter, mOutputLowPassFilter;
  90 public:
  91         int mBaseF0;
  92         float mBaseSpeed;
  93         float mBaseDeclination;
  94         int mBaseWaveform;
  95
  96         int mF0Flutter;
  97         int mSampleRate;
  98         int mNspFr;
  99         int mF0FundamentalFreq;        // Voicing fund freq in Hz
 100         int mVoicingAmpdb;          // Amp of voicing in dB,    0 to   70
 101         int mSkewnessOfAlternatePeriods;         // Skewness of alternate periods,0 to   40
 102         int mTimeCount;     // used for f0 flutter
 103         int mNPer;          // Current loc in voicing period   40000 samp/s
 104         int mT0;            // Fundamental period in output samples times 4
 105         int mNOpen;         // Number of samples in open phase of period
 106         int mNMod;          // Position in period to begin noise amp. modul
 107
 108         // Various amplitude variables used in main loop
 109
 110         float mAmpVoice;     // mVoicingAmpdb converted to linear gain
 111         float mAmpBypas;     // mBypassFricationAmpdb converted to linear gain
 112         float mAmpAspir;     // AP converted to linear gain
 113         float mAmpFrica;     // mFricationAmpdb converted to linear gain
 114         float mAmpBreth;     // ATURB converted to linear gain
 115
 116         // State variables of sound sources
 117
 118         int mSkew;                  // Alternating jitter, in half-period units
 119         float mVLast;               // Previous output of voice
 120         float mNLast;               // Previous output of random number generator
 121         float mGlotLast;            // Previous value of glotout
 122         float mDecay;               // mVoicingSpectralTiltdb converted to exponential time const
 123         float mOneMd;               // in voicing one-pole ELM_FEATURE_LOW-pass filter
 124
 125         unsigned int mSeed;                     // random seed
 126
 127
 128
 129         float natural_source(int aNper);
 130
 131         void frame_init();
 132         void flutter();
 133         void pitch_synch_par_reset(int ns);
 134         void parwave(short int *jwave);
 135         void init(int aBaseFrequency = 1330, float aBaseSpeed = 10.0f, float aBaseDeclination = 0.5f, int aBaseWaveform = KW_SAW);
 136         static int phone_to_elm(char *aPhoneme, int aCount, darray *aElement);
 137
 138         int mElementCount;
 139         unsigned char *mElement;
 140         int mElementIndex;
 141         klatt_frame mFrame;
 142         Element * mLastElement;
 143         int mTStress;
 144         int mNTStress;
 145         Slope mStressS;
 146         Slope mStressE;
 147         float mTop;
 148         void initsynth(int aElementCount,unsigned char *aElement);
 149         int synth(int aSampleCount, short *aSamplePointer);
 150         klatt();
 151 };
 152
 153 #endif