echo.h

Go to the documentation of this file.
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 ------------------------------------------------------------*/

Generated on Thu Dec 9 21:11:49 2010 for spandsp by  doxygen 1.5.9