Mercurial > hg > audiostuff

comparison spandsp-0.0.3/spandsp-0.0.3/src/spandsp/echo.h @ 5:f762bf195c4b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
import spandsp-0.0.3
author Peter Meerwald <pmeerw@cosy.sbg.ac.at>
date Fri, 25 Jun 2010 16:00:21 +0200
parents
children
comparison
equal deleted inserted replaced
4:26cd8f1ef0b1 5:f762bf195c4b
1 /*
2 * SpanDSP - a series of DSP components for telephony
3 *
4 * echo.c - A line echo canceller. This code is being developed
5 * against and partially complies with G168.
6 *
7 * Written by Steve Underwood <steveu@coppice.org>
8 * and David Rowe <david_at_rowetel_dot_com>
9 *
10 * Copyright (C) 2001 Steve Underwood and 2007 David Rowe
11 *
12 * All rights reserved.
13 *
14 * This program is free software; you can redistribute it and/or modify
15 * it under the terms of the GNU General Public License version 2, as
16 * published by the Free Software Foundation.
17 *
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
22 *
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
26 *
27 * $Id: echo.h,v 1.9 2006/10/24 13:45:28 steveu Exp $
28 */
29
30 /*! \file */
31
32 #if !defined(_ECHO_H_)
33 #define _ECHO_H_
34
35 /*! \page echo_can_page Line echo cancellation for voice
36
37 \section echo_can_page_sec_1 What does it do?
38 This module aims to provide G.168-2002 compliant echo cancellation, to remove
39 electrical echoes (e.g. from 2-4 wire hybrids) from voice calls.
40
41 \section echo_can_page_sec_2 How does it work?
42 The heart of the echo cancellor is FIR filter. This is adapted to match the echo
43 impulse response of the telephone line. It must be long enough to adequately cover
44 the duration of that impulse response. The signal transmitted to the telephone line
45 is passed through the FIR filter. Once the FIR is properly adapted, the resulting
46 output is an estimate of the echo signal received from the line. This is subtracted
47 from the received signal. The result is an estimate of the signal which originated
48 at the far end of the line, free from echos of our own transmitted signal.
49
50 The least mean squares (LMS) algorithm is attributed to Widrow and Hoff, and was
51 introduced in 1960. It is the commonest form of filter adaption used in things
52 like modem line equalisers and line echo cancellers. There it works very well.
53 However, it only works well for signals of constant amplitude. It works very poorly
54 for things like speech echo cancellation, where the signal level varies widely.
55 This is quite easy to fix. If the signal level is normalised - similar to applying
56 AGC - LMS can work as well for a signal of varying amplitude as it does for a modem
57 signal. This normalised least mean squares (NLMS) algorithm is the commonest one used
58 for speech echo cancellation. Many other algorithms exist - e.g. RLS (essentially
59 the same as Kalman filtering), FAP, etc. Some perform significantly better than NLMS.
60 However, factors such as computational complexity and patents favour the use of NLMS.
61
62 A simple refinement to NLMS can improve its performance with speech. NLMS tends
63 to adapt best to the strongest parts of a signal. If the signal is white noise,
64 the NLMS algorithm works very well. However, speech has more low frequency than
65 high frequency content. Pre-whitening (i.e. filtering the signal to flatten
66 its spectrum) the echo signal improves the adapt rate for speech, and ensures the
67 final residual signal is not heavily biased towards high frequencies. A very low
68 complexity filter is adequate for this, so pre-whitening adds little to the
69 compute requirements of the echo canceller.
70
71 An FIR filter adapted using pre-whitened NLMS performs well, provided certain
72 conditions are met:
73
74 - The transmitted signal has poor self-correlation.
75 - There is no signal being generated within the environment being cancelled.
76
77 The difficulty is that neither of these can be guaranteed.
78
79 If the adaption is performed while transmitting noise (or something fairly noise
80 like, such as voice) the adaption works very well. If the adaption is performed
81 while transmitting something highly correlative (typically narrow band energy
82 such as signalling tones or DTMF), the adaption can go seriously wrong. The reason
83 is there is only one solution for the adaption on a near random signal - the impulse
84 response of the line. For a repetitive signal, there are any number of solutions
85 which converge the adaption, and nothing guides the adaption to choose the generalised
86 one. Allowing an untrained canceller to converge on this kind of narrowband
87 energy probably a good thing, since at least it cancels the tones. Allowing a well
88 converged canceller to continue converging on such energy is just a way to ruin
89 its generalised adaption. A narrowband detector is needed, so adapation can be
90 suspended at appropriate times.
91
92 The adaption process is based on trying to eliminate the received signal. When
93 there is any signal from within the environment being cancelled it may upset the
94 adaption process. Similarly, if the signal we are transmitting is small, noise
95 may dominate and disturb the adaption process. If we can ensure that the
96 adaption is only performed when we are transmitting a significant signal level,
97 and the environment is not, things will be OK. Clearly, it is easy to tell when
98 we are sending a significant signal. Telling, if the environment is generating a
99 significant signal, and doing it with sufficient speed that the adaption will
100 not have diverged too much more we stop it, is a little harder.
101
102 The key problem in detecting when the environment is sourcing significant energy
103 is that we must do this very quickly. Given a reasonably long sample of the
104 received signal, there are a number of strategies which may be used to assess
105 whether that signal contains a strong far end component. However, by the time
106 that assessment is complete the far end signal will have already caused major
107 mis-convergence in the adaption process. An assessment algorithm is needed which
108 produces a fairly accurate result from a very short burst of far end energy.
109
110 \section echo_can_page_sec_3 How do I use it?
111 The echo cancellor processes both the transmit and receive streams sample by
112 sample. The processing function is not declared inline. Unfortunately,
113 cancellation requires many operations per sample, so the call overhead is only a
114 minor burden.
115 */
116
117 #include "fir.h"
118
119 /* Mask bits for the adaption mode */
120
121 #define ECHO_CAN_USE_ADAPTION 0x01
122 #define ECHO_CAN_USE_NLP 0x02
123 #define ECHO_CAN_USE_CNG 0x04
124 #define ECHO_CAN_USE_CLIP 0x08
125 #define ECHO_CAN_USE_TX_HPF 0x10
126 #define ECHO_CAN_USE_RX_HPF 0x20
127 #define ECHO_CAN_DISABLE 0x40
128
129 /*!
130 G.168 echo canceller descriptor. This defines the working state for a line
131 echo canceller.
132 */
133 typedef struct
134 {
135 int16_t tx,rx;
136 int16_t clean;
137 int16_t clean_nlp;
138
139 int nonupdate_dwell;
140 int curr_pos;
141 int taps;
142 int log2taps;
143 int adaption_mode;
144
145 int cond_met;
146 int32_t Pstates;
147 int16_t adapt;
148 int32_t factor;
149 int16_t shift;
150
151 /* Average levels and averaging filter states */
152 int Ltxacc, Lrxacc, Lcleanacc, Lclean_bgacc;
153 int Ltx, Lrx;
154 int Lclean;
155 int Lclean_bg;
156 int Lbgn, Lbgn_acc, Lbgn_upper, Lbgn_upper_acc;
157
158 /* foreground and background filter states */
159 fir16_state_t fir_state;
160 fir16_state_t fir_state_bg;
161 int16_t *fir_taps16[2];
162
163 /* DC blocking filter states */
164 int tx_1, tx_2, rx_1, rx_2;
165
166 /* optional High Pass Filter states */
167 int32_t xvtx[5], yvtx[5];
168 int32_t xvrx[5], yvrx[5];
169
170 /* Parameters for the optional Hoth noise generator */
171 int cng_level;
172 int cng_rndnum;
173 int cng_filter;
174
175 /* snapshot sample of coeffs used for development */
176 int16_t *snapshot;
177
178 } echo_can_state_t;
179
180 /*! Create a voice echo canceller context.
181 \param len The length of the canceller, in samples.
182 \return The new canceller context, or NULL if the canceller could not be created.
183 */
184 echo_can_state_t *echo_can_create(int len, int adaption_mode);
185
186 /*! Free a voice echo canceller context.
187 \param ec The echo canceller context.
188 */
189 void echo_can_free(echo_can_state_t *ec);
190
191 /*! Flush (reinitialise) a voice echo canceller context.
192 \param ec The echo canceller context.
193 */
194 void echo_can_flush(echo_can_state_t *ec);
195
196 /*! Set the adaption mode of a voice echo canceller context.
197 \param ec The echo canceller context.
198 \param adapt The mode.
199 */
200 void echo_can_adaption_mode(echo_can_state_t *ec, int adaption_mode);
201
202 void echo_can_snapshot(echo_can_state_t *ec);
203
204 /*! Process a sample through a voice echo canceller.
205 \param ec The echo canceller context.
206 \param tx The transmitted audio sample.
207 \param rx The received audio sample.
208 \return The clean (echo cancelled) received sample.
209 */
210 int16_t echo_can_update(echo_can_state_t *ec, int16_t tx, int16_t rx);
211
212 /*! Process to high pass filter the tx signal.
213 \param ec The echo canceller context.
214 \param tx The transmitted auio sample.
215 \return The HP filtered transmit sample, send this to your D/A.
216 */
217 int16_t echo_can_hpf_tx(echo_can_state_t *ec, int16_t tx);
218
219 #endif
220 /*- End of file ------------------------------------------------------------*/

Repositories maintained by Peter Meerwald, pmeerw@pmeerw.net.