1 
2 /*
3  * SpanDSP - a series of DSP components for telephony
4  *
5  * bit_operations.h - Various bit level operations, such as bit reversal
6  *
7  * Written by Steve Underwood <steveu@coppice.org>
8  *
9  * Copyright (C) 2006 Steve Underwood
10  *
11  * All rights reserved.
12  *
13  * This program is free software; you can redistribute it and/or modify
14  * it under the terms of the GNU General Public License version 2, as
15  * published by the Free Software Foundation.
16  *
17  * This program is distributed in the hope that it will be useful,
18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20  * GNU General Public License for more details.
21  *
22  * You should have received a copy of the GNU General Public License
23  * along with this program; if not, write to the Free Software
24  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
25  *
26  * $Id: bit_operations.h,v 1.15 2007/02/23 13:16:13 steveu Exp $
27  */
28 
29 /*! \file */
30 
31 #ifndef _CELLIAX_SPANDSP_H
32 #define _CELLIAX_SPANDSP_H
33 
34 #include <math.h>
35 
36 /*! \brief Find the bit position of the highest set bit in a word
37     \param bits The word to be searched
38     \return The bit number of the highest set bit, or -1 if the word is zero. */
top_bit(unsigned int bits)39 static __inline__ int top_bit(unsigned int bits)
40 {
41   int res;
42 
43 #if defined(__i386__)  ||  defined(__x86_64__)
44 __asm__(" xorl %[res],%[res];\n" " decl %[res];\n" " bsrl %[bits],%[res]\n":[res] "=&r"
45           (res)
46 :        [bits] "rm"(bits));
47   return res;
48 #elif defined(__ppc__)  ||   defined(__powerpc__)
49 __asm__("cntlzw %[res],%[bits];\n":[res] "=&r"(res)
50 :        [bits] "r"(bits));
51   return 31 - res;
52 #else
53   if (bits == 0)
54     return -1;
55   res = 0;
56   if (bits & 0xFFFF0000) {
57     bits &= 0xFFFF0000;
58     res += 16;
59   }
60   if (bits & 0xFF00FF00) {
61     bits &= 0xFF00FF00;
62     res += 8;
63   }
64   if (bits & 0xF0F0F0F0) {
65     bits &= 0xF0F0F0F0;
66     res += 4;
67   }
68   if (bits & 0xCCCCCCCC) {
69     bits &= 0xCCCCCCCC;
70     res += 2;
71   }
72   if (bits & 0xAAAAAAAA) {
73     bits &= 0xAAAAAAAA;
74     res += 1;
75   }
76   return res;
77 #endif
78 }
79 
80 /*- End of function --------------------------------------------------------*/
81 
82 /*! \brief Find the bit position of the lowest set bit in a word
83     \param bits The word to be searched
84     \return The bit number of the lowest set bit, or -1 if the word is zero. */
bottom_bit(unsigned int bits)85 static __inline__ int bottom_bit(unsigned int bits)
86 {
87   int res;
88 
89 #if defined(__i386__)  ||  defined(__x86_64__)
90 __asm__(" xorl %[res],%[res];\n" " decl %[res];\n" " bsfl %[bits],%[res]\n":[res] "=&r"
91           (res)
92 :        [bits] "rm"(bits));
93   return res;
94 #else
95   if (bits == 0)
96     return -1;
97   res = 31;
98   if (bits & 0x0000FFFF) {
99     bits &= 0x0000FFFF;
100     res -= 16;
101   }
102   if (bits & 0x00FF00FF) {
103     bits &= 0x00FF00FF;
104     res -= 8;
105   }
106   if (bits & 0x0F0F0F0F) {
107     bits &= 0x0F0F0F0F;
108     res -= 4;
109   }
110   if (bits & 0x33333333) {
111     bits &= 0x33333333;
112     res -= 2;
113   }
114   if (bits & 0x55555555) {
115     bits &= 0x55555555;
116     res -= 1;
117   }
118   return res;
119 #endif
120 }
121 
122 /*- End of function --------------------------------------------------------*/
123 
124 /*! \brief Bit reverse a byte.
125     \param data The byte to be reversed.
126     \return The bit reversed version of data. */
bit_reverse8(uint8_t x)127 static __inline__ uint8_t bit_reverse8(uint8_t x)
128 {
129 #if defined(__i386__)  ||  defined(__x86_64__)  ||  defined(__ppc__)  ||  defined(__powerpc__)
130   /* If multiply is fast */
131   return ((x * 0x0802U & 0x22110U) | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16;
132 #else
133   /* If multiply is slow, but we have a barrel shifter */
134   x = (x >> 4) | (x << 4);
135   x = ((x & 0xCC) >> 2) | ((x & 0x33) << 2);
136   return ((x & 0xAA) >> 1) | ((x & 0x55) << 1);
137 #endif
138 }
139 
140 /*- End of function --------------------------------------------------------*/
141 
142 /*! \brief Bit reverse a 16 bit word.
143     \param data The word to be reversed.
144     \return The bit reversed version of data. */
145 uint16_t bit_reverse16(uint16_t data);
146 
147 /*! \brief Bit reverse a 32 bit word.
148     \param data The word to be reversed.
149     \return The bit reversed version of data. */
150 uint32_t bit_reverse32(uint32_t data);
151 
152 /*! \brief Bit reverse each of the four bytes in a 32 bit word.
153     \param data The word to be reversed.
154     \return The bit reversed version of data. */
155 uint32_t bit_reverse_4bytes(uint32_t data);
156 
157 #if defined(__x86_64__)
158 /*! \brief Bit reverse each of the eight bytes in a 64 bit word.
159     \param data The word to be reversed.
160     \return The bit reversed version of data. */
161 uint64_t bit_reverse_8bytes(uint64_t data);
162 #endif
163 
164 /*! \brief Bit reverse each bytes in a buffer.
165     \param to The buffer to place the reversed data in.
166     \param from The buffer containing the data to be reversed.
167     \param The length of the data in the buffer. */
168 void bit_reverse(uint8_t to[], const uint8_t from[], int len);
169 
170 /*! \brief Find the number of set bits in a 32 bit word.
171     \param x The word to be searched.
172     \return The number of set bits. */
173 int one_bits32(uint32_t x);
174 
175 /*! \brief Create a mask as wide as the number in a 32 bit word.
176     \param x The word to be searched.
177     \return The mask. */
178 uint32_t make_mask32(uint32_t x);
179 
180 /*! \brief Create a mask as wide as the number in a 16 bit word.
181     \param x The word to be searched.
182     \return The mask. */
183 uint16_t make_mask16(uint16_t x);
184 
185 /*! \brief Find the least significant one in a word, and return a word
186            with just that bit set.
187     \param x The word to be searched.
188     \return The word with the single set bit. */
least_significant_one32(uint32_t x)189 static __inline__ uint32_t least_significant_one32(uint32_t x)
190 {
191   return (x & (-(int32_t) x));
192 }
193 
194 /*- End of function --------------------------------------------------------*/
195 
196 /*! \brief Find the most significant one in a word, and return a word
197            with just that bit set.
198     \param x The word to be searched.
199     \return The word with the single set bit. */
most_significant_one32(uint32_t x)200 static __inline__ uint32_t most_significant_one32(uint32_t x)
201 {
202 #if defined(__i386__)  ||  defined(__x86_64__)  ||  defined(__ppc__)  ||  defined(__powerpc__)
203   return 1 << top_bit(x);
204 #else
205   x = make_mask32(x);
206   return (x ^ (x >> 1));
207 #endif
208 }
209 
210 /*- End of function --------------------------------------------------------*/
211 
212 /*! \brief Find the parity of a byte.
213     \param x The byte to be checked.
214     \return 1 for odd, or 0 for even. */
parity8(uint8_t x)215 static __inline__ int parity8(uint8_t x)
216 {
217   x = (x ^ (x >> 4)) & 0x0F;
218   return (0x6996 >> x) & 1;
219 }
220 
221 /*- End of function --------------------------------------------------------*/
222 
223 /*! \brief Find the parity of a 16 bit word.
224     \param x The word to be checked.
225     \return 1 for odd, or 0 for even. */
parity16(uint16_t x)226 static __inline__ int parity16(uint16_t x)
227 {
228   x ^= (x >> 8);
229   x = (x ^ (x >> 4)) & 0x0F;
230   return (0x6996 >> x) & 1;
231 }
232 
233 /*- End of function --------------------------------------------------------*/
234 
235 /*! \brief Find the parity of a 32 bit word.
236     \param x The word to be checked.
237     \return 1 for odd, or 0 for even. */
parity32(uint32_t x)238 static __inline__ int parity32(uint32_t x)
239 {
240   x ^= (x >> 16);
241   x ^= (x >> 8);
242   x = (x ^ (x >> 4)) & 0x0F;
243   return (0x6996 >> x) & 1;
244 }
245 
246 /*- End of function --------------------------------------------------------*/
247 
248 /*- End of file ------------------------------------------------------------*/
249 /*
250  * SpanDSP - a series of DSP components for telephony
251  *
252  * fir.h - General telephony FIR routines
253  *
254  * Written by Steve Underwood <steveu@coppice.org>
255  *
256  * Copyright (C) 2002 Steve Underwood
257  *
258  * All rights reserved.
259  *
260  * This program is free software; you can redistribute it and/or modify
261  * it under the terms of the GNU General Public License version 2, as
262  * published by the Free Software Foundation.
263  *
264  * This program is distributed in the hope that it will be useful,
265  * but WITHOUT ANY WARRANTY; without even the implied warranty of
266  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
267  * GNU General Public License for more details.
268  *
269  * You should have received a copy of the GNU General Public License
270  * along with this program; if not, write to the Free Software
271  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
272  *
273  * $Id: fir.h,v 1.8 2006/10/24 13:45:28 steveu Exp $
274  */
275 
276 /*! \page fir_page FIR filtering
277 \section fir_page_sec_1 What does it do?
278 ???.
279 
280 \section fir_page_sec_2 How does it work?
281 ???.
282 */
283 
284 #if 0
285 #if defined(USE_MMX)  ||  defined(USE_SSE2)
286 #include "mmx.h"
287 #endif
288 
289 /*!
290     16 bit integer FIR descriptor. This defines the working state for a single
291     instance of an FIR filter using 16 bit integer coefficients.
292 */
293 typedef struct {
294   int taps;
295   int curr_pos;
296   const int16_t *coeffs;
297   int16_t *history;
298 } fir16_state_t;
299 
300 /*!
301     32 bit integer FIR descriptor. This defines the working state for a single
302     instance of an FIR filter using 32 bit integer coefficients, and filtering
303     16 bit integer data.
304 */
305 typedef struct {
306   int taps;
307   int curr_pos;
308   const int32_t *coeffs;
309   int16_t *history;
310 } fir32_state_t;
311 
312 /*!
313     Floating point FIR descriptor. This defines the working state for a single
314     instance of an FIR filter using floating point coefficients and data.
315 */
316 typedef struct {
317   int taps;
318   int curr_pos;
319   const float *coeffs;
320   float *history;
321 } fir_float_state_t;
322 
323 static __inline__ const int16_t *fir16_create(fir16_state_t * fir, const int16_t * coeffs,
324                                               int taps)
325 {
326   fir->taps = taps;
327   fir->curr_pos = taps - 1;
328   fir->coeffs = coeffs;
329 #if defined(USE_MMX)  ||  defined(USE_SSE2)
330   if ((fir->history = malloc(2 * taps * sizeof(int16_t))))
331     memset(fir->history, 0, 2 * taps * sizeof(int16_t));
332 #else
333   if ((fir->history = (int16_t *) malloc(taps * sizeof(int16_t))))
334     memset(fir->history, 0, taps * sizeof(int16_t));
335 #endif
336   return fir->history;
337 }
338 
339 /*- End of function --------------------------------------------------------*/
340 
341 static __inline__ void fir16_flush(fir16_state_t * fir)
342 {
343 #if defined(USE_MMX)  ||  defined(USE_SSE2)
344   memset(fir->history, 0, 2 * fir->taps * sizeof(int16_t));
345 #else
346   memset(fir->history, 0, fir->taps * sizeof(int16_t));
347 #endif
348 }
349 
350 /*- End of function --------------------------------------------------------*/
351 
352 static __inline__ void fir16_free(fir16_state_t * fir)
353 {
354   free(fir->history);
355 }
356 
357 /*- End of function --------------------------------------------------------*/
358 
359 static __inline__ int16_t fir16(fir16_state_t * fir, int16_t sample)
360 {
361   int i;
362   int32_t y;
363 #if defined(USE_MMX)
364   mmx_t *mmx_coeffs;
365   mmx_t *mmx_hist;
366 
367   fir->history[fir->curr_pos] = sample;
368   fir->history[fir->curr_pos + fir->taps] = sample;
369 
370   mmx_coeffs = (mmx_t *) fir->coeffs;
371   mmx_hist = (mmx_t *) & fir->history[fir->curr_pos];
372   i = fir->taps;
373   pxor_r2r(mm4, mm4);
374   /* 8 samples per iteration, so the filter must be a multiple of 8 long. */
375   while (i > 0) {
376     movq_m2r(mmx_coeffs[0], mm0);
377     movq_m2r(mmx_coeffs[1], mm2);
378     movq_m2r(mmx_hist[0], mm1);
379     movq_m2r(mmx_hist[1], mm3);
380     mmx_coeffs += 2;
381     mmx_hist += 2;
382     pmaddwd_r2r(mm1, mm0);
383     pmaddwd_r2r(mm3, mm2);
384     paddd_r2r(mm0, mm4);
385     paddd_r2r(mm2, mm4);
386     i -= 8;
387   }
388   movq_r2r(mm4, mm0);
389   psrlq_i2r(32, mm0);
390   paddd_r2r(mm0, mm4);
391   movd_r2m(mm4, y);
392   emms();
393 #elif defined(USE_SSE2)
394   xmm_t *xmm_coeffs;
395   xmm_t *xmm_hist;
396 
397   fir->history[fir->curr_pos] = sample;
398   fir->history[fir->curr_pos + fir->taps] = sample;
399 
400   xmm_coeffs = (xmm_t *) fir->coeffs;
401   xmm_hist = (xmm_t *) & fir->history[fir->curr_pos];
402   i = fir->taps;
403   pxor_r2r(xmm4, xmm4);
404   /* 16 samples per iteration, so the filter must be a multiple of 16 long. */
405   while (i > 0) {
406     movdqu_m2r(xmm_coeffs[0], xmm0);
407     movdqu_m2r(xmm_coeffs[1], xmm2);
408     movdqu_m2r(xmm_hist[0], xmm1);
409     movdqu_m2r(xmm_hist[1], xmm3);
410     xmm_coeffs += 2;
411     xmm_hist += 2;
412     pmaddwd_r2r(xmm1, xmm0);
413     pmaddwd_r2r(xmm3, xmm2);
414     paddd_r2r(xmm0, xmm4);
415     paddd_r2r(xmm2, xmm4);
416     i -= 16;
417   }
418   movdqa_r2r(xmm4, xmm0);
419   psrldq_i2r(8, xmm0);
420   paddd_r2r(xmm0, xmm4);
421   movdqa_r2r(xmm4, xmm0);
422   psrldq_i2r(4, xmm0);
423   paddd_r2r(xmm0, xmm4);
424   movd_r2m(xmm4, y);
425 #else
426   int offset1;
427   int offset2;
428 
429   fir->history[fir->curr_pos] = sample;
430 
431   offset2 = fir->curr_pos;
432   offset1 = fir->taps - offset2;
433   y = 0;
434   for (i = fir->taps - 1; i >= offset1; i--)
435     y += fir->coeffs[i] * fir->history[i - offset1];
436   for (; i >= 0; i--)
437     y += fir->coeffs[i] * fir->history[i + offset2];
438 #endif
439   if (fir->curr_pos <= 0)
440     fir->curr_pos = fir->taps;
441   fir->curr_pos--;
442   return (int16_t) (y >> 15);
443 }
444 
445 /*- End of function --------------------------------------------------------*/
446 
447 static __inline__ const int16_t *fir32_create(fir32_state_t * fir, const int32_t * coeffs,
448                                               int taps)
449 {
450   fir->taps = taps;
451   fir->curr_pos = taps - 1;
452   fir->coeffs = coeffs;
453   fir->history = (int16_t *) malloc(taps * sizeof(int16_t));
454   if (fir->history)
455     memset(fir->history, '\0', taps * sizeof(int16_t));
456   return fir->history;
457 }
458 
459 /*- End of function --------------------------------------------------------*/
460 
461 static __inline__ void fir32_flush(fir32_state_t * fir)
462 {
463   memset(fir->history, 0, fir->taps * sizeof(int16_t));
464 }
465 
466 /*- End of function --------------------------------------------------------*/
467 
468 static __inline__ void fir32_free(fir32_state_t * fir)
469 {
470   free(fir->history);
471 }
472 
473 /*- End of function --------------------------------------------------------*/
474 
475 static __inline__ int16_t fir32(fir32_state_t * fir, int16_t sample)
476 {
477   int i;
478   int32_t y;
479   int offset1;
480   int offset2;
481 
482   fir->history[fir->curr_pos] = sample;
483   offset2 = fir->curr_pos;
484   offset1 = fir->taps - offset2;
485   y = 0;
486   for (i = fir->taps - 1; i >= offset1; i--)
487     y += fir->coeffs[i] * fir->history[i - offset1];
488   for (; i >= 0; i--)
489     y += fir->coeffs[i] * fir->history[i + offset2];
490   if (fir->curr_pos <= 0)
491     fir->curr_pos = fir->taps;
492   fir->curr_pos--;
493   return (int16_t) (y >> 15);
494 }
495 
496 /*- End of function --------------------------------------------------------*/
497 
498 static __inline__ const float *fir_float_create(fir_float_state_t * fir,
499                                                 const float *coeffs, int taps)
500 {
501   fir->taps = taps;
502   fir->curr_pos = taps - 1;
503   fir->coeffs = coeffs;
504   fir->history = (float *) malloc(taps * sizeof(float));
505   if (fir->history)
506     memset(fir->history, '\0', taps * sizeof(float));
507   return fir->history;
508 }
509 
510 /*- End of function --------------------------------------------------------*/
511 
512 static __inline__ void fir_float_free(fir_float_state_t * fir)
513 {
514   free(fir->history);
515 }
516 
517 /*- End of function --------------------------------------------------------*/
518 
519 static __inline__ int16_t fir_float(fir_float_state_t * fir, int16_t sample)
520 {
521   int i;
522   float y;
523   int offset1;
524   int offset2;
525 
526   fir->history[fir->curr_pos] = sample;
527 
528   offset2 = fir->curr_pos;
529   offset1 = fir->taps - offset2;
530   y = 0;
531   for (i = fir->taps - 1; i >= offset1; i--)
532     y += fir->coeffs[i] * fir->history[i - offset1];
533   for (; i >= 0; i--)
534     y += fir->coeffs[i] * fir->history[i + offset2];
535   if (fir->curr_pos <= 0)
536     fir->curr_pos = fir->taps;
537   fir->curr_pos--;
538   return (int16_t) y;
539 }
540 
541 /*- End of function --------------------------------------------------------*/
542 #endif
543 
544 /*- End of file ------------------------------------------------------------*/
545 
546 /*
547  * SpanDSP - a series of DSP components for telephony
548  *
549  * echo.h - An echo cancellor, suitable for electrical and acoustic
550  *	    cancellation. This code does not currently comply with
551  *	    any relevant standards (e.g. G.164/5/7/8).
552  *
553  * Written by Steve Underwood <steveu@coppice.org>
554  *
555  * Copyright (C) 2001 Steve Underwood
556  *
557  * Based on a bit from here, a bit from there, eye of toad,
558  * ear of bat, etc - plus, of course, my own 2 cents.
559  *
560  * All rights reserved.
561  *
562  * This program is free software; you can redistribute it and/or modify
563  * it under the terms of the GNU General Public License version 2, as
564  * published by the Free Software Foundation.
565  *
566  * This program is distributed in the hope that it will be useful,
567  * but WITHOUT ANY WARRANTY; without even the implied warranty of
568  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
569  * GNU General Public License for more details.
570  *
571  * You should have received a copy of the GNU General Public License
572  * along with this program; if not, write to the Free Software
573  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
574  *
575  * $Id: echo.h,v 1.9 2006/10/24 13:45:28 steveu Exp $
576  */
577 
578 /*! \file */
579 
580 /*! \page echo_can_page Line echo cancellation for voice
581 
582 \section echo_can_page_sec_1 What does it do?
583 This module aims to provide G.168-2002 compliant echo cancellation, to remove
584 electrical echoes (e.g. from 2-4 wire hybrids) from voice calls.
585 
586 \section echo_can_page_sec_2 How does it work?
587 The heart of the echo cancellor is FIR filter. This is adapted to match the echo
588 impulse response of the telephone line. It must be long enough to adequately cover
589 the duration of that impulse response. The signal transmitted to the telephone line
590 is passed through the FIR filter. Once the FIR is properly adapted, the resulting
591 output is an estimate of the echo signal received from the line. This is subtracted
592 from the received signal. The result is an estimate of the signal which originated
593 at the far end of the line, free from echos of our own transmitted signal.
594 
595 The least mean squares (LMS) algorithm is attributed to Widrow and Hoff, and was
596 introduced in 1960. It is the commonest form of filter adaption used in things
597 like modem line equalisers and line echo cancellers. There it works very well.
598 However, it only works well for signals of constant amplitude. It works very poorly
599 for things like speech echo cancellation, where the signal level varies widely.
600 This is quite easy to fix. If the signal level is normalised - similar to applying
601 AGC - LMS can work as well for a signal of varying amplitude as it does for a modem
602 signal. This normalised least mean squares (NLMS) algorithm is the commonest one used
603 for speech echo cancellation. Many other algorithms exist - e.g. RLS (essentially
604 the same as Kalman filtering), FAP, etc. Some perform significantly better than NLMS.
605 However, factors such as computational complexity and patents favour the use of NLMS.
606 
607 A simple refinement to NLMS can improve its performance with speech. NLMS tends
608 to adapt best to the strongest parts of a signal. If the signal is white noise,
609 the NLMS algorithm works very well. However, speech has more low frequency than
610 high frequency content. Pre-whitening (i.e. filtering the signal to flatten
611 its spectrum) the echo signal improves the adapt rate for speech, and ensures the
612 final residual signal is not heavily biased towards high frequencies. A very low
613 complexity filter is adequate for this, so pre-whitening adds little to the
614 compute requirements of the echo canceller.
615 
616 An FIR filter adapted using pre-whitened NLMS performs well, provided certain
617 conditions are met:
618 
619     - The transmitted signal has poor self-correlation.
620     - There is no signal being generated within the environment being cancelled.
621 
622 The difficulty is that neither of these can be guaranteed.
623 
624 If the adaption is performed while transmitting noise (or something fairly noise
625 like, such as voice) the adaption works very well. If the adaption is performed
626 while transmitting something highly correlative (typically narrow band energy
627 such as signalling tones or DTMF), the adaption can go seriously wrong. The reason
628 is there is only one solution for the adaption on a near random signal - the impulse
629 response of the line. For a repetitive signal, there are any number of solutions
630 which converge the adaption, and nothing guides the adaption to choose the generalised
631 one. Allowing an untrained canceller to converge on this kind of narrowband
632 energy probably a good thing, since at least it cancels the tones. Allowing a well
633 converged canceller to continue converging on such energy is just a way to ruin
634 its generalised adaption. A narrowband detector is needed, so adapation can be
635 suspended at appropriate times.
636 
637 The adaption process is based on trying to eliminate the received signal. When
638 there is any signal from within the environment being cancelled it may upset the
639 adaption process. Similarly, if the signal we are transmitting is small, noise
640 may dominate and disturb the adaption process. If we can ensure that the
641 adaption is only performed when we are transmitting a significant signal level,
642 and the environment is not, things will be OK. Clearly, it is easy to tell when
643 we are sending a significant signal. Telling, if the environment is generating a
644 significant signal, and doing it with sufficient speed that the adaption will
645 not have diverged too much more we stop it, is a little harder.
646 
647 The key problem in detecting when the environment is sourcing significant energy
648 is that we must do this very quickly. Given a reasonably long sample of the
649 received signal, there are a number of strategies which may be used to assess
650 whether that signal contains a strong far end component. However, by the time
651 that assessment is complete the far end signal will have already caused major
652 mis-convergence in the adaption process. An assessment algorithm is needed which
653 produces a fairly accurate result from a very short burst of far end energy.
654 
655 \section echo_can_page_sec_3 How do I use it?
656 The echo cancellor processes both the transmit and receive streams sample by
657 sample. The processing function is not declared inline. Unfortunately,
658 cancellation requires many operations per sample, so the call overhead is only a
659 minor burden.
660 */
661 
662 #define NONUPDATE_DWELL_TIME	600 /* 600 samples, or 75ms */
663 
664 #if 0
665 /* Mask bits for the adaption mode */
666 #define ECHO_CAN_USE_NLP            0x01
667 #define ECHO_CAN_USE_SUPPRESSOR     0x02
668 #define ECHO_CAN_USE_CNG            0x04
669 #define ECHO_CAN_USE_ADAPTION       0x08
670 
671 /*!
672     G.168 echo canceller descriptor. This defines the working state for a line
673     echo canceller.
674 */
675 typedef struct {
676   int tx_power[4];
677   int rx_power[3];
678   int clean_rx_power;
679 
680   int rx_power_threshold;
681   int nonupdate_dwell;
682 
683   fir16_state_t fir_state;
684   /*! Echo FIR taps (16 bit version) */
685   int16_t *fir_taps16[4];
686   /*! Echo FIR taps (32 bit version) */
687   int32_t *fir_taps32;
688 
689   int curr_pos;
690 
691   int taps;
692   int tap_mask;
693   int adaption_mode;
694 
695   int32_t supp_test1;
696   int32_t supp_test2;
697   int32_t supp1;
698   int32_t supp2;
699   int vad;
700   int cng;
701   /* Parameters for the Hoth noise generator */
702   int cng_level;
703   int cng_rndnum;
704   int cng_filter;
705 
706   int16_t geigel_max;
707   int geigel_lag;
708   int dtd_onset;
709   int tap_set;
710   int tap_rotate_counter;
711 
712   int32_t latest_correction;    /* Indication of the magnitude of the latest
713                                    adaption, or a code to indicate why adaption
714                                    was skipped, for test purposes */
715   int32_t last_acf[28];
716   int narrowband_count;
717   int narrowband_score;
718 } echo_can_state_t;
719 
720 /*! Create a voice echo canceller context.
721     \param len The length of the canceller, in samples.
722     \return The new canceller context, or NULL if the canceller could not be created.
723 */
724 echo_can_state_t *echo_can_create(int len, int adaption_mode);
725 
726 /*! Free a voice echo canceller context.
727     \param ec The echo canceller context.
728 */
729 void echo_can_free(echo_can_state_t * ec);
730 
731 /*! Flush (reinitialise) a voice echo canceller context.
732     \param ec The echo canceller context.
733 */
734 void echo_can_flush(echo_can_state_t * ec);
735 
736 /*! Set the adaption mode of a voice echo canceller context.
737     \param ec The echo canceller context.
738     \param adapt The mode.
739 */
740 void echo_can_adaption_mode(echo_can_state_t * ec, int adaption_mode);
741 
742 /*! Process a sample through a voice echo canceller.
743     \param ec The echo canceller context.
744     \param tx The transmitted audio sample.
745     \param rx The received audio sample.
746     \return The clean (echo cancelled) received sample.
747 */
748 int16_t echo_can_update(echo_can_state_t * ec, int16_t tx, int16_t rx);
749 
750 #endif
751 /*- End of file ------------------------------------------------------------*/
752 
753 /*!
754     Floating point Goertzel filter descriptor.
755 */
756 typedef struct {
757   float fac;
758   int samples;
759 } goertzel_descriptor_t;
760 
761 /*!
762     Floating point Goertzel filter state descriptor.
763 */
764 typedef struct {
765   float v2;
766   float v3;
767   float fac;
768   int samples;
769   int current_sample;
770 } goertzel_state_t;
771 
772 /*! \brief Create a descriptor for use with either a Goertzel transform */
773 void make_goertzel_descriptor(goertzel_descriptor_t * t, float freq, int samples);
774 
775 /*! \brief Initialise the state of a Goertzel transform.
776     \param s The Goertzel context. If NULL, a context is allocated with malloc.
777     \param t The Goertzel descriptor.
778     \return A pointer to the Goertzel state. */
779 goertzel_state_t *goertzel_init(goertzel_state_t * s, goertzel_descriptor_t * t);
780 
781 /*! \brief Reset the state of a Goertzel transform.
782     \param s The Goertzel context.
783     \param t The Goertzel descriptor.
784     \return A pointer to the Goertzel state. */
785 void goertzel_reset(goertzel_state_t * s);
786 
787 /*! \brief Update the state of a Goertzel transform.
788     \param s The Goertzel context
789     \param amp The samples to be transformed
790     \param samples The number of samples
791     \return The number of samples unprocessed */
792 int goertzel_update(goertzel_state_t * s, const int16_t amp[], int samples);
793 
794 /*! \brief Evaluate the final result of a Goertzel transform.
795     \param s The Goertzel context
796     \return The result of the transform. */
797 float goertzel_result(goertzel_state_t * s);
798 
799 /*! \brief Update the state of a Goertzel transform.
800     \param s The Goertzel context
801     \param amp The sample to be transformed. */
goertzel_sample(goertzel_state_t * s,int16_t amp)802 static __inline__ void goertzel_sample(goertzel_state_t * s, int16_t amp)
803 {
804   float v1;
805 
806   v1 = s->v2;
807   s->v2 = s->v3;
808   s->v3 = s->fac * s->v2 - v1 + amp;
809   s->current_sample++;
810 }
811 
812 /*- End of function --------------------------------------------------------*/
813 
814 /*
815  * SpanDSP - a series of DSP components for telephony
816  *
817  * tone_detect.c - General telephony tone detection.
818  *
819  * Written by Steve Underwood <steveu@coppice.org>
820  *
821  * Copyright (C) 2001-2003, 2005 Steve Underwood
822  *
823  * All rights reserved.
824  *
825  * This program is free software; you can redistribute it and/or modify
826  * it under the terms of the GNU General Public License version 2, as
827  * published by the Free Software Foundation.
828  *
829  * This program is distributed in the hope that it will be useful,
830  * but WITHOUT ANY WARRANTY; without even the implied warranty of
831  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
832  * GNU General Public License for more details.
833  *
834  * You should have received a copy of the GNU General Public License
835  * along with this program; if not, write to the Free Software
836  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
837  *
838  * $Id: tone_detect.c,v 1.31 2007/03/03 10:40:33 steveu Exp $
839  */
840 
841 /*! \file tone_detect.h */
842 
843 #if !defined(M_PI)
844 /* C99 systems may not define M_PI */
845 #define M_PI 3.14159265358979323846264338327
846 #endif
847 /*! \page dtmf_rx_page DTMF receiver
848 \section dtmf_rx_page_sec_1 What does it do?
849 The DTMF receiver detects the standard DTMF digits. It is compliant with
850 ITU-T Q.23, ITU-T Q.24, and the local DTMF specifications of most administrations.
851 Its passes the test suites. It also scores *very* well on the standard
852 talk-off tests.
853 
854 The current design uses floating point extensively. It is not tolerant of DC.
855 It is expected that a DC restore stage will be placed before the DTMF detector.
856 Unless the dial tone filter is switched on, the detector has poor tolerance
857 of dial tone. Whether this matter depends on your application. If you are using
858 the detector in an IVR application you will need proper echo cancellation to
859 get good performance in the presence of speech prompts, so dial tone will not
860 exist. If you do need good dial tone tolerance, a dial tone filter can be
861 enabled in the detector.
862 
863 \section dtmf_rx_page_sec_2 How does it work?
864 Like most other DSP based DTMF detector's, this one uses the Goertzel algorithm
865 to look for the DTMF tones. What makes each detector design different is just how
866 that algorithm is used.
867 
868 Basic DTMF specs:
869     - Minimum tone on = 40ms
870     - Minimum tone off = 50ms
871     - Maximum digit rate = 10 per second
872     - Normal twist <= 8dB accepted
873     - Reverse twist <= 4dB accepted
874     - S/N >= 15dB will detect OK
875     - Attenuation <= 26dB will detect OK
876     - Frequency tolerance +- 1.5% will detect, +-3.5% will reject
877 
878 TODO:
879 */
880 
881 /*! \page dtmf_tx_page DTMF tone generation
882 \section dtmf_tx_page_sec_1 What does it do?
883 
884 The DTMF tone generation module provides for the generation of the
885 repertoire of 16 DTMF dual tones.
886 
887 \section dtmf_tx_page_sec_2 How does it work?
888 */
889 
890 #define MAX_DTMF_DIGITS 128
891 
892 typedef void (*dtmf_rx_callback_t) (void *user_data, const char *digits, int len);
893 
894 /*!
895     DTMF generator state descriptor. This defines the state of a single
896     working instance of a DTMF generator.
897 */
898 #if 0
899 typedef struct {
900   tone_gen_descriptor_t *tone_descriptors;
901   tone_gen_state_t tones;
902   char digits[MAX_DTMF_DIGITS + 1];
903   int current_sample;
904   size_t current_digits;
905 } dtmf_tx_state_t;
906 
907 #endif
908 
909 /*!
910     DTMF digit detector descriptor.
911 */
912 typedef struct {
913   /*! Optional callback funcion to deliver received digits. */
914   dtmf_rx_callback_t callback;
915   /*! An opaque pointer passed to the callback function. */
916   void *callback_data;
917   /*! Optional callback funcion to deliver real time digit state changes. */
918   //tone_report_func_t realtime_callback;
919   void *realtime_callback;
920   /*! An opaque pointer passed to the real time callback function. */
921   void *realtime_callback_data;
922   /*! TRUE if dialtone should be filtered before processing */
923   int filter_dialtone;
924   /*! Maximum acceptable "normal" (lower bigger than higher) twist ratio */
925   float normal_twist;
926   /*! Maximum acceptable "reverse" (higher bigger than lower) twist ratio */
927   float reverse_twist;
928 
929   /*! 350Hz filter state for the optional dialtone filter */
930   float z350_1;
931   float z350_2;
932   /*! 440Hz filter state for the optional dialtone filter */
933   float z440_1;
934   float z440_2;
935 
936   /*! Tone detector working states */
937   goertzel_state_t row_out[4];
938   goertzel_state_t col_out[4];
939   /*! The accumlating total energy on the same period over which the Goertzels work. */
940   float energy;
941   /*! The result of the last tone analysis. */
942   uint8_t last_hit;
943   /*! The confirmed digit we are currently receiving */
944   uint8_t in_digit;
945   /*! The current sample number within a processing block. */
946   int current_sample;
947 
948   /*! The received digits buffer. This is a NULL terminated string. */
949   char digits[MAX_DTMF_DIGITS + 1];
950   /*! The number of digits currently in the digit buffer. */
951   int current_digits;
952   /*! The number of digits which have been lost due to buffer overflows. */
953   int lost_digits;
954 } dtmf_rx_state_t;
955 
956 #if 0
957 /*! \brief Generate a buffer of DTMF tones.
958     \param s The DTMF generator context.
959     \param amp The buffer for the generated signal.
960     \param max_samples The required number of generated samples.
961     \return The number of samples actually generated. This may be less than
962             samples if the input buffer empties. */
963 int dtmf_tx(dtmf_tx_state_t * s, int16_t amp[], int max_samples);
964 
965 /*! \brief Put a string of digits in a DTMF generator's input buffer.
966     \param s The DTMF generator context.
967     \param digits The string of digits to be added.
968     \return The number of digits actually added. This may be less than the
969             length of the digit string, if the buffer fills up. */
970 size_t dtmf_tx_put(dtmf_tx_state_t * s, const char *digits);
971 
972 /*! \brief Initialise a DTMF tone generator context.
973     \param s The DTMF generator context.
974     \return A pointer to the DTMF generator context. */
975 dtmf_tx_state_t *dtmf_tx_init(dtmf_tx_state_t * s);
976 #endif
977 
978 /*! Set a optional realtime callback for a DTMF receiver context. This function
979     is called immediately a confirmed state change occurs in the received DTMF. It
980     is called with the ASCII value for a DTMF tone pair, or zero to indicate no tone
981     is being received.
982     \brief Set a realtime callback for a DTMF receiver context.
983     \param s The DTMF receiver context.
984     \param callback Callback routine used to report the start and end of digits.
985     \param user_data An opaque pointer which is associated with the context,
986            and supplied in callbacks. */
987 void dtmf_rx_set_realtime_callback(dtmf_rx_state_t * s,
988                                    //tone_report_func_t callback,
989                                    void *callback, void *user_data);
990 
991 /*! \brief Adjust a DTMF receiver context.
992     \param s The DTMF receiver context.
993     \param filter_dialtone TRUE to enable filtering of dialtone, FALSE
994            to disable, < 0 to leave unchanged.
995     \param twist Acceptable twist, in dB. < 0 to leave unchanged.
996     \param reverse_twist Acceptable reverse twist, in dB. < 0 to leave unchanged. */
997 void dtmf_rx_parms(dtmf_rx_state_t * s, int filter_dialtone, int twist,
998                    int reverse_twist);
999 
1000 /*! Process a block of received DTMF audio samples.
1001     \brief Process a block of received DTMF audio samples.
1002     \param s The DTMF receiver context.
1003     \param amp The audio sample buffer.
1004     \param samples The number of samples in the buffer.
1005     \return The number of samples unprocessed. */
1006 int dtmf_rx(dtmf_rx_state_t * s, const int16_t amp[], int samples);
1007 
1008 /*! \brief Get a string of digits from a DTMF receiver's output buffer.
1009     \param s The DTMF receiver context.
1010     \param digits The buffer for the received digits.
1011     \param max The maximum  number of digits to be returned,
1012     \return The number of digits actually returned. */
1013 size_t dtmf_rx_get(dtmf_rx_state_t * s, char *digits, int max);
1014 
1015 /*! \brief Initialise a DTMF receiver context.
1016     \param s The DTMF receiver context.
1017     \param callback An optional callback routine, used to report received digits. If
1018            no callback routine is set, digits may be collected, using the dtmf_rx_get()
1019            function.
1020     \param user_data An opaque pointer which is associated with the context,
1021            and supplied in callbacks.
1022     \return A pointer to the DTMF receiver context. */
1023 dtmf_rx_state_t *dtmf_rx_init(dtmf_rx_state_t * s, dtmf_rx_callback_t callback,
1024                               void *user_data);
1025 
1026 /*- End of file ------------------------------------------------------------*/
1027 
1028 #endif /* _CELLIAX_SPANDSP_H */
1029