spandsp
0.0.6
|
00001 /* 00002 * SpanDSP - a series of DSP components for telephony 00003 * 00004 * plc.h 00005 * 00006 * Written by Steve Underwood <steveu@coppice.org> 00007 * 00008 * Copyright (C) 2004 Steve Underwood 00009 * 00010 * All rights reserved. 00011 * 00012 * This program is free software; you can redistribute it and/or modify 00013 * it under the terms of the GNU Lesser General Public License version 2.1, 00014 * as published by the Free Software Foundation. 00015 * 00016 * This program is distributed in the hope that it will be useful, 00017 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00018 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 00019 * GNU Lesser General Public License for more details. 00020 * 00021 * You should have received a copy of the GNU Lesser General Public 00022 * License along with this program; if not, write to the Free Software 00023 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. 00024 */ 00025 00026 /*! \file */ 00027 00028 #if !defined(_SPANDSP_PLC_H_) 00029 #define _SPANDSP_PLC_H_ 00030 00031 /*! \page plc_page Packet loss concealment 00032 \section plc_page_sec_1 What does it do? 00033 The packet loss concealment module provides a synthetic fill-in signal, to minimise 00034 the audible effect of lost packets in VoIP applications. It is not tied to any 00035 particular codec, and could be used with almost any codec which does not 00036 specify its own procedure for packet loss concealment. 00037 00038 Where a codec specific concealment procedure exists, that algorithm is usually built 00039 around knowledge of the characteristics of the particular codec. It will, therefore, 00040 generally give better results for that particular codec than this generic concealer will. 00041 00042 The PLC code implements an algorithm similar to the one described in Appendix 1 of G.711. 00043 However, the G.711 algorithm is optimised for 10ms packets. Few people use such small 00044 packets. 20ms is a much more common value, and longer packets are also quite common. The 00045 algorithm has been adjusted with this in mind. Also, the G.711 approach causes an 00046 algorithmic delay, and requires significant buffer manipulation when there is no packet 00047 loss. The algorithm used here avoids this. It causes no delay, and achieves comparable 00048 quality with normal speech. 00049 00050 Note that both this algorithm, and the one in G.711 are optimised for speech. For most kinds 00051 of music a much slower decay on bursts of lost packets give better results. 00052 00053 \section plc_page_sec_2 How does it work? 00054 While good packets are being received, the plc_rx() routine keeps a record of the trailing 00055 section of the known speech signal. If a packet is missed, plc_fillin() is called to produce 00056 a synthetic replacement for the real speech signal. The average mean difference function 00057 (AMDF) is applied to the last known good signal, to determine its effective pitch. 00058 Based on this, the last pitch period of signal is saved. Essentially, this cycle of speech 00059 will be repeated over and over until the real speech resumes. However, several refinements 00060 are needed to obtain smooth pleasant sounding results. 00061 00062 - The two ends of the stored cycle of speech will not always fit together smoothly. This can 00063 cause roughness, or even clicks, at the joins between cycles. To soften this, the 00064 1/4 pitch period of real speech preceeding the cycle to be repeated is blended with the last 00065 1/4 pitch period of the cycle to be repeated, using an overlap-add (OLA) technique (i.e. 00066 in total, the last 5/4 pitch periods of real speech are used). 00067 00068 - The start of the synthetic speech will not always fit together smoothly with the tail of 00069 real speech passed on before the erasure was identified. Ideally, we would like to modify 00070 the last 1/4 pitch period of the real speech, to blend it into the synthetic speech. However, 00071 it is too late for that. We could have delayed the real speech a little, but that would 00072 require more buffer manipulation, and hurt the efficiency of the no-lost-packets case 00073 (which we hope is the dominant case). Instead we use a degenerate form of OLA to modify 00074 the start of the synthetic data. The last 1/4 pitch period of real speech is time reversed, 00075 and OLA is used to blend it with the first 1/4 pitch period of synthetic speech. The result 00076 seems quite acceptable. 00077 00078 - As we progress into the erasure, the chances of the synthetic signal being anything like 00079 correct steadily fall. Therefore, the volume of the synthesized signal is made to decay 00080 linearly, such that after 50ms of missing audio it is reduced to silence. 00081 00082 - When real speech resumes, an extra 1/4 pitch period of synthetic speech is blended with the 00083 start of the real speech. If the erasure is small, this smoothes the transition. If the erasure 00084 is long, and the synthetic signal has faded to zero, the blending softens the start up of the 00085 real signal, avoiding a kind of "click" or "pop" effect that might occur with a sudden onset. 00086 00087 \section plc_page_sec_3 How do I use it? 00088 Before audio is processed, call plc_init() to create an instance of the packet loss 00089 concealer. For each received audio packet that is acceptable (i.e. not including those being 00090 dropped for being too late) call plc_rx() to record the content of the packet. Note this may 00091 modify the packet a little after a period of packet loss, to blend real synthetic data smoothly. 00092 When a real packet is not available in time, call plc_fillin() to create a sythetic substitute. 00093 That's it! 00094 */ 00095 00096 /*! Minimum allowed pitch (66 Hz) */ 00097 #define PLC_PITCH_MIN 120 00098 /*! Maximum allowed pitch (200 Hz) */ 00099 #define PLC_PITCH_MAX 40 00100 /*! Maximum pitch OLA window */ 00101 #define PLC_PITCH_OVERLAP_MAX (PLC_PITCH_MIN >> 2) 00102 /*! The length over which the AMDF function looks for similarity (20 ms) */ 00103 #define CORRELATION_SPAN 160 00104 /*! History buffer length. The buffer much also be at leat 1.25 times 00105 PLC_PITCH_MIN, but that is much smaller than the buffer needs to be for 00106 the pitch assessment. */ 00107 #define PLC_HISTORY_LEN (CORRELATION_SPAN + PLC_PITCH_MIN) 00108 00109 /*! 00110 The generic packet loss concealer context. 00111 */ 00112 typedef struct 00113 { 00114 /*! Consecutive erased samples */ 00115 int missing_samples; 00116 /*! Current offset into pitch period */ 00117 int pitch_offset; 00118 /*! Pitch estimate */ 00119 int pitch; 00120 /*! Buffer for a cycle of speech */ 00121 float pitchbuf[PLC_PITCH_MIN]; 00122 /*! History buffer */ 00123 int16_t history[PLC_HISTORY_LEN]; 00124 /*! Current pointer into the history buffer */ 00125 int buf_ptr; 00126 } plc_state_t; 00127 00128 00129 #if defined(__cplusplus) 00130 extern "C" 00131 { 00132 #endif 00133 00134 /*! Process a block of received audio samples for PLC. 00135 \brief Process a block of received audio samples for PLC. 00136 \param s The packet loss concealer context. 00137 \param amp The audio sample buffer. 00138 \param len The number of samples in the buffer. 00139 \return The number of samples in the buffer. */ 00140 SPAN_DECLARE(int) plc_rx(plc_state_t *s, int16_t amp[], int len); 00141 00142 /*! Fill-in a block of missing audio samples. 00143 \brief Fill-in a block of missing audio samples. 00144 \param s The packet loss concealer context. 00145 \param amp The audio sample buffer. 00146 \param len The number of samples to be synthesised. 00147 \return The number of samples synthesized. */ 00148 SPAN_DECLARE(int) plc_fillin(plc_state_t *s, int16_t amp[], int len); 00149 00150 /*! Initialise a packet loss concealer context. 00151 \brief Initialise a PLC context. 00152 \param s The packet loss concealer context. 00153 \return A pointer to the the packet loss concealer context. */ 00154 SPAN_DECLARE(plc_state_t *) plc_init(plc_state_t *s); 00155 00156 /*! Release a packet loss concealer context. 00157 \param s The packet loss concealer context. 00158 \return 0 for OK. */ 00159 SPAN_DECLARE(int) plc_release(plc_state_t *s); 00160 00161 /*! Free a packet loss concealer context. 00162 \param s The packet loss concealer context. 00163 \return 0 for OK. */ 00164 SPAN_DECLARE(int) plc_free(plc_state_t *s); 00165 00166 #if defined(__cplusplus) 00167 } 00168 #endif 00169 00170 #endif 00171 /*- End of file ------------------------------------------------------------*/