spandsp
0.0.6
|
00001 /* 00002 * SpanDSP - a series of DSP components for telephony 00003 * 00004 * echo.h - An echo cancellor, suitable for electrical and acoustic 00005 * cancellation. This code does not currently comply with 00006 * any relevant standards (e.g. G.164/5/7/8). 00007 * 00008 * Written by Steve Underwood <steveu@coppice.org> 00009 * 00010 * Copyright (C) 2001 Steve Underwood 00011 * 00012 * All rights reserved. 00013 * 00014 * This program is free software; you can redistribute it and/or modify 00015 * it under the terms of the GNU Lesser General Public License version 2.1, 00016 * as published by the Free Software Foundation. 00017 * 00018 * This program is distributed in the hope that it will be useful, 00019 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00020 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 00021 * GNU Lesser General Public License for more details. 00022 * 00023 * You should have received a copy of the GNU Lesser General Public 00024 * License along with this program; if not, write to the Free Software 00025 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. 00026 */ 00027 00028 /*! \file */ 00029 00030 #if !defined(_SPANDSP_ECHO_H_) 00031 #define _SPANDSP_ECHO_H_ 00032 00033 /*! \page echo_can_page Line echo cancellation for voice 00034 00035 \section echo_can_page_sec_1 What does it do? 00036 This module aims to provide G.168-2002 compliant echo cancellation, to remove 00037 electrical echoes (e.g. from 2-4 wire hybrids) from voice calls. 00038 00039 \section echo_can_page_sec_2 How does it work? 00040 The heart of the echo cancellor is FIR filter. This is adapted to match the echo 00041 impulse response of the telephone line. It must be long enough to adequately cover 00042 the duration of that impulse response. The signal transmitted to the telephone line 00043 is passed through the FIR filter. Once the FIR is properly adapted, the resulting 00044 output is an estimate of the echo signal received from the line. This is subtracted 00045 from the received signal. The result is an estimate of the signal which originated 00046 at the far end of the line, free from echos of our own transmitted signal. 00047 00048 The least mean squares (LMS) algorithm is attributed to Widrow and Hoff, and was 00049 introduced in 1960. It is the commonest form of filter adaption used in things 00050 like modem line equalisers and line echo cancellers. There it works very well. 00051 However, it only works well for signals of constant amplitude. It works very poorly 00052 for things like speech echo cancellation, where the signal level varies widely. 00053 This is quite easy to fix. If the signal level is normalised - similar to applying 00054 AGC - LMS can work as well for a signal of varying amplitude as it does for a modem 00055 signal. This normalised least mean squares (NLMS) algorithm is the commonest one used 00056 for speech echo cancellation. Many other algorithms exist - e.g. RLS (essentially 00057 the same as Kalman filtering), FAP, etc. Some perform significantly better than NLMS. 00058 However, factors such as computational complexity and patents favour the use of NLMS. 00059 00060 A simple refinement to NLMS can improve its performance with speech. NLMS tends 00061 to adapt best to the strongest parts of a signal. If the signal is white noise, 00062 the NLMS algorithm works very well. However, speech has more low frequency than 00063 high frequency content. Pre-whitening (i.e. filtering the signal to flatten 00064 its spectrum) the echo signal improves the adapt rate for speech, and ensures the 00065 final residual signal is not heavily biased towards high frequencies. A very low 00066 complexity filter is adequate for this, so pre-whitening adds little to the 00067 compute requirements of the echo canceller. 00068 00069 An FIR filter adapted using pre-whitened NLMS performs well, provided certain 00070 conditions are met: 00071 00072 - The transmitted signal has poor self-correlation. 00073 - There is no signal being generated within the environment being cancelled. 00074 00075 The difficulty is that neither of these can be guaranteed. 00076 00077 If the adaption is performed while transmitting noise (or something fairly noise 00078 like, such as voice) the adaption works very well. If the adaption is performed 00079 while transmitting something highly correlative (typically narrow band energy 00080 such as signalling tones or DTMF), the adaption can go seriously wrong. The reason 00081 is there is only one solution for the adaption on a near random signal - the impulse 00082 response of the line. For a repetitive signal, there are any number of solutions 00083 which converge the adaption, and nothing guides the adaption to choose the generalised 00084 one. Allowing an untrained canceller to converge on this kind of narrowband 00085 energy probably a good thing, since at least it cancels the tones. Allowing a well 00086 converged canceller to continue converging on such energy is just a way to ruin 00087 its generalised adaption. A narrowband detector is needed, so adapation can be 00088 suspended at appropriate times. 00089 00090 The adaption process is based on trying to eliminate the received signal. When 00091 there is any signal from within the environment being cancelled it may upset the 00092 adaption process. Similarly, if the signal we are transmitting is small, noise 00093 may dominate and disturb the adaption process. If we can ensure that the 00094 adaption is only performed when we are transmitting a significant signal level, 00095 and the environment is not, things will be OK. Clearly, it is easy to tell when 00096 we are sending a significant signal. Telling, if the environment is generating a 00097 significant signal, and doing it with sufficient speed that the adaption will 00098 not have diverged too much more we stop it, is a little harder. 00099 00100 The key problem in detecting when the environment is sourcing significant energy 00101 is that we must do this very quickly. Given a reasonably long sample of the 00102 received signal, there are a number of strategies which may be used to assess 00103 whether that signal contains a strong far end component. However, by the time 00104 that assessment is complete the far end signal will have already caused major 00105 mis-convergence in the adaption process. An assessment algorithm is needed which 00106 produces a fairly accurate result from a very short burst of far end energy. 00107 00108 \section echo_can_page_sec_3 How do I use it? 00109 The echo cancellor processes both the transmit and receive streams sample by 00110 sample. The processing function is not declared inline. Unfortunately, 00111 cancellation requires many operations per sample, so the call overhead is only a 00112 minor burden. 00113 */ 00114 00115 #include "fir.h" 00116 00117 /* Mask bits for the adaption mode */ 00118 enum 00119 { 00120 ECHO_CAN_USE_ADAPTION = 0x01, 00121 ECHO_CAN_USE_NLP = 0x02, 00122 ECHO_CAN_USE_CNG = 0x04, 00123 ECHO_CAN_USE_CLIP = 0x08, 00124 ECHO_CAN_USE_SUPPRESSOR = 0x10, 00125 ECHO_CAN_USE_TX_HPF = 0x20, 00126 ECHO_CAN_USE_RX_HPF = 0x40, 00127 ECHO_CAN_DISABLE = 0x80 00128 }; 00129 00130 /*! 00131 G.168 echo canceller descriptor. This defines the working state for a line 00132 echo canceller. 00133 */ 00134 typedef struct echo_can_state_s echo_can_state_t; 00135 00136 #if defined(__cplusplus) 00137 extern "C" 00138 { 00139 #endif 00140 00141 /*! Create a voice echo canceller context. 00142 \param len The length of the canceller, in samples. 00143 \return The new canceller context, or NULL if the canceller could not be created. 00144 */ 00145 SPAN_DECLARE(echo_can_state_t *) echo_can_init(int len, int adaption_mode); 00146 00147 /*! Release a voice echo canceller context. 00148 \param ec The echo canceller context. 00149 \return 0 for OK, else -1. 00150 */ 00151 SPAN_DECLARE(int) echo_can_release(echo_can_state_t *ec); 00152 00153 /*! Free a voice echo canceller context. 00154 \param ec The echo canceller context. 00155 \return 0 for OK, else -1. 00156 */ 00157 SPAN_DECLARE(int) echo_can_free(echo_can_state_t *ec); 00158 00159 /*! Flush (reinitialise) a voice echo canceller context. 00160 \param ec The echo canceller context. 00161 */ 00162 SPAN_DECLARE(void) echo_can_flush(echo_can_state_t *ec); 00163 00164 /*! Set the adaption mode of a voice echo canceller context. 00165 \param ec The echo canceller context. 00166 \param adaption_mode The mode. 00167 */ 00168 SPAN_DECLARE(void) echo_can_adaption_mode(echo_can_state_t *ec, int adaption_mode); 00169 00170 /*! Process a sample through a voice echo canceller. 00171 \param ec The echo canceller context. 00172 \param tx The transmitted audio sample. 00173 \param rx The received audio sample. 00174 \return The clean (echo cancelled) received sample. 00175 */ 00176 SPAN_DECLARE(int16_t) echo_can_update(echo_can_state_t *ec, int16_t tx, int16_t rx); 00177 00178 /*! Process to high pass filter the tx signal. 00179 \param ec The echo canceller context. 00180 \param tx The transmitted auio sample. 00181 \return The HP filtered transmit sample, send this to your D/A. 00182 */ 00183 SPAN_DECLARE(int16_t) echo_can_hpf_tx(echo_can_state_t *ec, int16_t tx); 00184 00185 SPAN_DECLARE(void) echo_can_snapshot(echo_can_state_t *ec); 00186 00187 #if defined(__cplusplus) 00188 } 00189 #endif 00190 00191 #endif 00192 /*- End of file ------------------------------------------------------------*/