/************************************************************************\   Low bitrate LPC CODEC derived from the public domain implementation  of Ron Frederick.   The basic design is preserved, except for several bug fixes and  the following modifications:   1. The pitch detector operates on the (downsampled) signal, not on  the residue. This appears to yield better performances, and it  lowers the processing load.  2. Each frame is elongated by 50% prefixing it with the last half  of the previous frame. This design, inherited from the original  code for windowing purposes, is exploited in order to provide  two independent pitch analyses: on the first 2/3, and on the  second 2/3 of the elongated frame (of course, they overlap by  half):   last half old frame	            new frame  --------------------========================================  <--------- first pitch region --------->                      <--------- second pitch region  ------->   Two voiced/unvoiced flags define the voicing status of each  region; only one value for the period is retained (if both  halves are voiced, the average is used).  The two flags are used by the synthesizer for the halves of  each frame to play back. Of course, this is non optimal but  is good enough (a half-frame would be too short for measuring  low pitches)  3. The parameters (one float for the period (pitch), one for the  gain, and ten for the LPC-10 filter) are quantized according  this procedure:  - the period is logarithmically compressed, then quantized  as 8-bit unsigned int (6 would actually suffice)  - the gain is logarithmically compressed (using a different  formula), then quantized at 6-bit unsigned int. The two  remaining bits are used for the voicing flags.  - the first two LPC parameters (k[1] and k[2]) are multiplied  by PI/2, and the arcsine of the result is quantized as  6 and 5 bit signed integers. This has proved more effective  than the log-area compression used by LPC-10.  - the remaining eight LPC parameters (k[3]...k[10]) are  quantized as, respectively, 5,4,4,3,3,3,3 and 2 bit signed  integers.  Finally, period and gain plus voicing flags are stored in the  first two bytes of the 7-byte parameters block, and the quantized  LPC parameters are packed into the remaining 5 bytes. Two bits  remain unassigned, and can be used for error detection or other  purposes.   The frame lenght is actually variable, and is simply passed as  initialization parameter to lpc_init(): this allows to experiment  with various frame lengths. Long frames reduce the bitrate, but  exceeding 320 samples (i.e. 40 ms, at 8000 samples/s) tend to  deteriorate the speech, that sounds like spoken by a person  affected by a stroke: the LPC parameters (modeling the vocal  tract) can't change fast enough for a natural-sounding synthesis.  25 ms per frame already yields a quite tight compression, corresponding  to 1000/40 * 7 * 8 = 1400 bps. The quality improves little with  frames shorter than 250 samples (32 frames/s), so this is a recommended  compromise. The bitrate is 32 * 7 * 8 = 1792 bps.   The synthesizer has been modified as well. For voiced subframes it  now uses a sawtooth excitation, instead of the original pulse train.  This idea, copied from MELP, reduces the buzzing-noise artifacts.  In order to compensate the non-white spectrum of the sawtooth, a  pre-emphasis is applied to the signal before the Durbin calculation.  The filter has (in s-space) two zeroes at (640, 0) Hz and two poles  at (3200, 0) Hz. These filters have been handoded, and may not be  optimal. Two other filters (anti-hum high-pass with corner at 100 Hz,  and pre-downsampling lowpass with corner at 300 Hz) are Butterworth  designs produced by the MkFilter package by A.J. Fisher  (http://www.cs.york.ac.uk/~fisher/mkfilter/).   The C style has been ANSI-fied.   Complexity: As any LPC CODEC, also this one is not very demanding:  for real-time use analysis and synthesis takes each about 6 - 8%  of the CPU cycles on a Cy686/166, when the code is compiled with  MSVC++ 4.2 with /Ox or gcc with -O3.  However, a floating point unit is absolutely required.  \************************************************************************/ #include <stdio.h>#include <math.h>#include <string.h>#include <stdlib.h> #include "soundpipe.h"#include "openlpc.h"#ifndef __plan9__#include "ftol.h"#endif #define PREEMPH #define bcopy(a, b, n)	  memmove(b, a, n) #ifndef M_PI#define M_PI (3.1415926535897932384626433832795)#endif static float my_fs = 11025.0;#define LPC_FILTORDER		10#define FS		my_fs /* Sampling rate */#define MAXWINDOW	1000	/* Max analysis window length */ typedef struct openlpc_e_state{	float   s[MAXWINDOW], y[MAXWINDOW], h[MAXWINDOW];    int     framelen, buflen;    float   xv1[3], yv1[3],            xv2[2], yv2[2],			xv3[1], yv3[3],			xv4[2], yv4[2];    float   w[MAXWINDOW], r[LPC_FILTORDER+1];} openlpc_e_state_t; typedef struct openlpc_d_state{		float Oldper, OldG, Oldk[LPC_FILTORDER + 1];        float bp[LPC_FILTORDER+1];        float exc;		int pitchctr, framelen, buflen;} openlpc_d_state_t; #define FC		200.0	/* Pitch analyzer filter cutoff */#define DOWN		5	/* Decimation for pitch analyzer */#define MINPIT		40.0	/* Minimum pitch (observed: 74) */#define MAXPIT		320.0	/* Maximum pitch (observed: 250) */ #define MINPER		7 //(int)(FS/(DOWN*MAXPIT)+.5)	/* Minimum period  */#define MAXPER		55 //(int)(FS/(DOWN*MINPIT)+.5)	/* Maximum period  */ #define REAL_MINPER	 (DOWN*MINPER) /* converted to samples units */ #define WSCALE		1.5863	/* Energy loss due to windowing */ #define BITS_FOR_LPC 38 #define ARCSIN_Q /* provides better quantization of first two k[] at low bitrates */ #if BITS_FOR_LPC == 38/* (38 bit LPC-10, 2.7 Kbit/s @ 20ms, 2.4 Kbit/s @ 22.5 ms */static int parambits[LPC_FILTORDER] = {6,5,5,4,4,3,3,3,3,2};#elif BITS_FOR_LPC == 32/* (32 bit LPC-10, 2.4 Kbit/s, not so good */static int parambits[LPC_FILTORDER] = {5,5,5,4,3,3,2,2,2,1};#else /* BITS_FOR_LPC == 80	*//* 80-bit LPC10, 4.8 Kbit/s */static int parambits[LPC_FILTORDER] = {8,8,8,8,8,8,8,8,8,8};#endif static float logmaxminper;static int sizeofparm;	/* computed by lpc_init */ static void auto_correl1(float *w, int n, float *r){    int i, k;     for (k=0; k <= MAXPER; k++, n--) {        r[k] = 0.0;        for (i=0; i < n; i++) {            r[k] += (w[i] *  w[i+k]);        }    }} static void auto_correl2(float *w, int n, float *r){    int i, k;     for (k=0; k <= LPC_FILTORDER; k++, n--) {        r[k] = 0.0;        for (i=0; i < n; i++) {            r[k] += (w[i] *  w[i+k]);        }    }} static void durbin(float r[], int p, float k[], float *g){    int i, j;    float a[LPC_FILTORDER+1], at[LPC_FILTORDER+1], e;     for (i=0; i <= p; i++) a[i] = at[i] = 0.0;     e = r[0];    for (i=1; i <= p; i++) {        k[i] = -r[i];        for (j=1; j < i; j++) {            at[j] = a[j];            k[i] -= a[j] * r[i-j];        }        if (e == 0) {  /* fix by John Walker */            *g = 0;            return;        }        k[i] /= e;        a[i] = k[i];        for (j=1; j < i; j++) a[j] = at[j] + k[i] * at[i-j];        e *= 1.0f - k[i]*k[i];    }    if (e < 0) {        e = 0; /* fix by John Walker */    }    *g = (float)sqrt(e);} /* Enzo's streamlined pitch extractor - on the signal, not the residue */ static void calc_pitch(float *w, int len, float *per){    int i, j, rpos;    float d[MAXWINDOW/DOWN];    float r[MAXPER+1];    float rval, rm, rp, rmax;    float x, y;    float vthresh;     /* decimation */    for (i=0, j=0; i < len; i+=DOWN)        d[j++] = w[i];     auto_correl1(d, len/DOWN, r);     /* find peak between MINPER and MAXPER */    x = 1;    rpos = 0;    rmax = 0.0;     for (i = 1; i < MAXPER; i++) {        rm = r[i-1];        rp = r[i+1];        y = rm+r[i]+rp; /* find max of integral from i-1 to i+1 */         if ((y > rmax) && (r[i] > rm) && (r[i] > rp) && (i > MINPER)) {            rmax = y;            rpos = i;        }    }     /* consider adjacent values */    rm = r[rpos-1];    rp = r[rpos+1]; #if 0    {        float a, b, c, x, y;        /* parabolic interpolation */        a = 0.5f * rm - rmax + 0.5f * rp;        b = -0.5f*rm*(2.0f*rpos+1.0f) + 2.0f*rpos*rmax + 0.5f*rp*(1.0f-2.0f*rpos);        c = 0.5f*rm*(rpos*rpos+rpos) + rmax*(1.0f-rpos*rpos) + 0.5f*rp*(rpos*rpos-rpos);         /* find max of interpolating parabole */        x = -b / (2.0f * a);        y = a*x*x + b*x + c;         rmax = y;        /* normalize, so that 0. < rval < 1. */        rval = (r[0] == 0 ? 1.0f : rmax / r[0]);    }#else    if(rpos > 0) {        x = ((rpos-1)*rm + rpos*r[rpos] + (rpos+1)*rp)/(rm+r[rpos]+rp);    }    /* normalize, so that 0. < rval < 1. */    rval = (r[0] == 0 ? 0 : r[rpos] / r[0]);#endif     /* periods near the low boundary and at low volumes    are usually spurious and    manifest themselves as annoying mosquito buzzes */     *per = 0;	/* default: unvoiced */    if ( x > MINPER &&  /* x could be < MINPER or even < 0 if pos == MINPER */        x < (MAXPER+1) /* same story */        ) {         vthresh = 0.6f;        if(r[0] > 0.002)	   /* at low volumes (< 0.002), prefer unvoiced */            vthresh = 0.25;       /* drop threshold at high volumes */         if(rval > vthresh)            *per = x * DOWN;    }} /* Initialization of various parameters */ openlpc_encoder_state *create_openlpc_encoder_state(void){    openlpc_encoder_state *state;     state = (openlpc_encoder_state *)calloc(1, sizeof(openlpc_encoder_state));     return state;}  void init_openlpc_encoder_state(openlpc_encoder_state *st, int framelen){    int i, j;     st->framelen = framelen;     st->buflen = framelen*3/2;    /*  (st->buflen > MAXWINDOW) return -1;*/     for(i=0, j=0; i<sizeof(parambits)/sizeof(parambits[0]); i++) {        j += parambits[i];    }    sizeofparm = (j+7)/8 + 2;    for (i = 0; i < st->buflen; i++) {        st->s[i] = 0.0;        st->h[i] = (float)(WSCALE*(0.54 - 0.46 * cos(2 * M_PI * i / (st->buflen-1.0))));    }    /* init the filters */    st->xv1[0] = st->xv1[1] = st->xv1[2] = st->yv1[0] = st->yv1[1] = st->yv1[2] = 0.0f;    st->xv2[0] = st->xv2[1] = st->yv2[0] = st->yv2[1] = 0.0f;    st->xv3[0] = st->yv3[0] = st->yv3[1] = st->yv3[2] = 0.0f;    st->xv4[0] = st->xv4[1] = st->yv4[0] = st->yv4[1] = 0.0f;     logmaxminper = (float)log((float)MAXPER/MINPER); } void destroy_openlpc_encoder_state(openlpc_encoder_state *st){    free(st);} /* LPC Analysis (compression) */ int openlpc_encode(const short *buf, unsigned char *parm, openlpc_encoder_state *st){    int i, j;    float per, gain, k[LPC_FILTORDER+1];    float per1, per2;    float xv10, xv11, xv12, yv10, yv11, yv12,        xv20, xv21, yv20, yv21,        xv30, yv30, yv31, yv32,        xv40, xv41, yv40, yv41;     xv10 = st->xv1[0];    xv11 = st->xv1[1];    xv12 = st->xv1[2];    yv10 = st->yv1[0];    yv11 = st->yv1[1];    yv12 = st->yv1[2];    xv30 = st->xv3[0];    yv30 = st->yv3[0];    yv31 = st->yv3[1];    yv32 = st->yv3[2];    for(i = 0; i < LPC_FILTORDER + 1; i++) k[i] = 0;    /* convert short data in buf[] to signed lin. data in s[] and prefilter */    for (i=0, j=st->buflen - st->framelen; i < st->framelen; i++, j++) {         float u = (float)(buf[i]/32768.0f);         /* Anti-hum 2nd order Butterworth high-pass, 100 Hz corner frequency */        /* Digital filter designed by mkfilter/mkshape/gencode   A.J. Fisher        mkfilter -Bu -Hp -o 2 -a 0.0125 -l -z */         xv10 = xv11;        xv11 = xv12;        xv12 = (float)(u * 0.94597831f); /* /GAIN */         yv10 = yv11;        yv11 = yv12;        yv12 = (float)((xv10 + xv12) - 2 * xv11            + ( -0.8948742499f * yv10) + ( 1.8890389823f * yv11));         u = st->s[j] = yv12;	/* also affects input of next stage, to the LPC filter synth */         /* low-pass filter s[] -> y[] before computing pitch */        /* second-order Butterworth low-pass filter, corner at 300 Hz */        /* Digital filter designed by mkfilter/mkshape/gencode   A.J. Fisher        MKFILTER.EXE -Bu -Lp -o 2 -a 0.0375 -l -z */         /*st->xv3[0] = (float)(u / 2.127814584e+001);*/ /* GAIN */        xv30 = (float)(u * 0.04699658f); /* GAIN */        yv30 = yv31;        yv31 = yv32;        yv32 = xv30 + (float)(( -0.7166152306f * yv30) + (1.6696186545f * yv31));        st->y[j] = yv32;    }    st->xv1[0] = xv10;    st->xv1[1] = xv11;    st->xv1[2] = xv12;    st->yv1[0] = yv10;    st->yv1[1] = yv11;    st->yv1[2] = yv12;    st->xv3[0] = xv30;    st->yv3[0] = yv30;    st->yv3[1] = yv31;    st->yv3[2] = yv32;#ifdef PREEMPH    /* operate optional preemphasis s[] -> s[] on the newly arrived frame */    xv20 = st->xv2[0];    xv21 = st->xv2[1];    yv20 = st->yv2[0];    yv21 = st->yv2[1];    xv40 = st->xv4[0];    xv41 = st->xv4[1];    yv40 = st->yv4[0];    yv41 = st->yv4[1];    for (j=st->buflen - st->framelen; j < st->buflen; j++) {        float u = st->s[j];         /* handcoded filter: 1 zero at 640 Hz, 1 pole at 3200 */#define TAU (FS/3200.f)#define RHO (0.1f)        xv20 = xv21; 	/* e(n-1) */        xv21 = (float)(u * 1.584f);		/* e(n)	, add 4 dB to compensate attenuation */        yv20 = yv21;        yv21 = (float)(TAU/(1.f+RHO+TAU) * yv20 	 /* u(n) */            + (RHO+TAU)/(1.f+RHO+TAU) * xv21            - TAU/(1.f+RHO+TAU) * xv20);        u = yv21;         /* cascaded copy of handcoded filter: 1 zero at 640 Hz, 1 pole at 3200 */        xv40 = xv41;        xv41 = (float)(u * 1.584f);        yv40 = yv41;        yv41 = (float)(TAU/(1.f+RHO+TAU) * yv40            + (RHO+TAU)/(1.f+RHO+TAU) * xv41            - TAU/(1.f+RHO+TAU) * xv40);        u = yv41;         st->s[j] = u;    }    st->xv2[0] = xv20;    st->xv2[1] = xv21;    st->yv2[0] = yv20;    st->yv2[1] = yv21;    st->xv4[0] = xv40;    st->xv4[1] = xv41;    st->yv4[0] = yv40;    st->yv4[1] = yv41;#endif     /* operate windowing s[] -> w[] */     for (i=0; i < st->buflen; i++)        st->w[i] = st->s[i] * st->h[i];     /* compute LPC coeff. from autocorrelation (first 11 values) of windowed data */    auto_correl2(st->w, st->buflen, st->r);    durbin(st->r, LPC_FILTORDER, k, &gain);     /* calculate pitch */    calc_pitch(st->y, st->framelen, &per1);                 /* first 2/3 of buffer */    calc_pitch(st->y + st->buflen - st->framelen, st->framelen, &per2); /* last 2/3 of buffer */    if(per1 > 0 && per2 >0)        per = (per1+per2)/2;    else if(per1 > 0)        per = per1;    else if(per2 > 0)        per = per2;    else        per = 0;     /* logarithmic q.: 0 = MINPER, 256 = MAXPER */    parm[0] = (unsigned char)(per == 0? 0 : (unsigned char)(log(per/(REAL_MINPER)) / logmaxminper * (1<<8))); #ifdef LINEAR_G_Q    i = gain * (1<<7);    if(i > 255) 	/* bug fix by EM */        i = 255;#else    i = (int)(float)(256.0f * log(1 + (2.718-1.f)/10.f*gain)); /* deriv = 5.82 allowing to reserve 2 bits */    if(i > 255) i = 255;	 /* reached when gain = 10 */    i = (i+2) & 0xfc;#endif     parm[1] = (unsigned char)i;     if(per1 > 0)        parm[1] |= 1;    if(per2 > 0)        parm[1] |= 2;     for(j=2; j < sizeofparm; j++)        parm[j] = 0;     for (i=0; i < LPC_FILTORDER; i++) {        int bitamount = parambits[i];        int bitc8 = 8-bitamount;        int q = (1 << bitc8);  /* quantum: 1, 2, 4... */        float u = k[i+1];        int iu;#ifdef ARCSIN_Q        if(i < 2) u = (float)(asin(u)*2.f/M_PI);#endif        u *= 127;        if(u < 0)            u += (0.6f * q);        else

Lines 1–500 of 756. Load more