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
 28#ifndef __ECHO_H
 29#define __ECHO_H
 30
 31/*
 32Line echo cancellation for voice
 33
 34What does it do?
 35
 36This module aims to provide G.168-2002 compliant echo cancellation, to remove
 37electrical echoes (e.g. from 2-4 wire hybrids) from voice calls.
 38
 39
 40How does it work?
 41
 42The heart of the echo cancellor is FIR filter. This is adapted to match the
 43echo impulse response of the telephone line. It must be long enough to
 44adequately cover the duration of that impulse response. The signal transmitted
 45to the telephone line is passed through the FIR filter. Once the FIR is
 46properly adapted, the resulting output is an estimate of the echo signal
 47received from the line. This is subtracted from the received signal. The result
 48is an estimate of the signal which originated at the far end of the line, free
 49from echos of our own transmitted signal.
 50
 51The least mean squares (LMS) algorithm is attributed to Widrow and Hoff, and
 52was introduced in 1960. It is the commonest form of filter adaption used in
 53things like modem line equalisers and line echo cancellers. There it works very
 54well.  However, it only works well for signals of constant amplitude. It works
 55very poorly for things like speech echo cancellation, where the signal level
 56varies widely.  This is quite easy to fix. If the signal level is normalised -
 57similar to applying AGC - LMS can work as well for a signal of varying
 58amplitude as it does for a modem signal. This normalised least mean squares
 59(NLMS) algorithm is the commonest one used for speech echo cancellation. Many
 60other algorithms exist - e.g. RLS (essentially the same as Kalman filtering),
 61FAP, etc. Some perform significantly better than NLMS.  However, factors such
 62as computational complexity and patents favour the use of NLMS.
 63
 64A simple refinement to NLMS can improve its performance with speech. NLMS tends
 65to adapt best to the strongest parts of a signal. If the signal is white noise,
 66the NLMS algorithm works very well. However, speech has more low frequency than
 67high frequency content. Pre-whitening (i.e. filtering the signal to flatten its
 68spectrum) the echo signal improves the adapt rate for speech, and ensures the
 69final residual signal is not heavily biased towards high frequencies. A very
 70low complexity filter is adequate for this, so pre-whitening adds little to the
 71compute requirements of the echo canceller.
 72
 73An FIR filter adapted using pre-whitened NLMS performs well, provided certain
 74conditions are met:
 75
 76    - The transmitted signal has poor self-correlation.
 77    - There is no signal being generated within the environment being
 78      cancelled.
 79
 80The difficulty is that neither of these can be guaranteed.
 81
 82If the adaption is performed while transmitting noise (or something fairly
 83noise like, such as voice) the adaption works very well. If the adaption is
 84performed while transmitting something highly correlative (typically narrow
 85band energy such as signalling tones or DTMF), the adaption can go seriously
 86wrong. The reason is there is only one solution for the adaption on a near
 87random signal - the impulse response of the line. For a repetitive signal,
 88there are any number of solutions which converge the adaption, and nothing
 89guides the adaption to choose the generalised one. Allowing an untrained
 90canceller to converge on this kind of narrowband energy probably a good thing,
 91since at least it cancels the tones. Allowing a well converged canceller to
 92continue converging on such energy is just a way to ruin its generalised
 93adaption. A narrowband detector is needed, so adapation can be suspended at
 94appropriate times.
 95
 96The adaption process is based on trying to eliminate the received signal. When
 97there is any signal from within the environment being cancelled it may upset
 98the adaption process. Similarly, if the signal we are transmitting is small,
 99noise may dominate and disturb the adaption process. If we can ensure that the
100adaption is only performed when we are transmitting a significant signal level,
101and the environment is not, things will be OK. Clearly, it is easy to tell when
102we are sending a significant signal. Telling, if the environment is generating
103a significant signal, and doing it with sufficient speed that the adaption will
104not have diverged too much more we stop it, is a little harder.
105
106The key problem in detecting when the environment is sourcing significant
107energy is that we must do this very quickly. Given a reasonably long sample of
108the received signal, there are a number of strategies which may be used to
109assess whether that signal contains a strong far end component. However, by the
110time that assessment is complete the far end signal will have already caused
111major mis-convergence in the adaption process. An assessment algorithm is
112needed which produces a fairly accurate result from a very short burst of far
113end energy.
114
115How do I use it?
116
117The echo cancellor processes both the transmit and receive streams sample by
118sample. The processing function is not declared inline. Unfortunately,
119cancellation requires many operations per sample, so the call overhead is only
120a minor burden.
121*/
122
123#include "fir.h"
124#include "oslec.h"
125
126/*
127    G.168 echo canceller descriptor. This defines the working state for a line
128    echo canceller.
129*/
130struct oslec_state {
131	int16_t tx, rx;
132	int16_t clean;
133	int16_t clean_nlp;
134
135	int nonupdate_dwell;
136	int curr_pos;
137	int taps;
138	int log2taps;
139	int adaption_mode;
140
141	int cond_met;
142	int32_t Pstates;
143	int16_t adapt;
144	int32_t factor;
145	int16_t shift;
146
147	/* Average levels and averaging filter states */
148	int Ltxacc, Lrxacc, Lcleanacc, Lclean_bgacc;
149	int Ltx, Lrx;
150	int Lclean;
151	int Lclean_bg;
152	int Lbgn, Lbgn_acc, Lbgn_upper, Lbgn_upper_acc;
153
154	/* foreground and background filter states */
155	struct fir16_state_t fir_state;
156	struct fir16_state_t fir_state_bg;
157	int16_t *fir_taps16[2];
158
159	/* DC blocking filter states */
160	int tx_1, tx_2, rx_1, rx_2;
161
162	/* optional High Pass Filter states */
163	int32_t xvtx[5], yvtx[5];
164	int32_t xvrx[5], yvrx[5];
165
166	/* Parameters for the optional Hoth noise generator */
167	int cng_level;
168	int cng_rndnum;
169	int cng_filter;
170
171	/* snapshot sample of coeffs used for development */
172	int16_t *snapshot;
173};
174
175#endif /* __ECHO_H */