2 * audio encoder psychoacoustic model
3 * Copyright (C) 2008 Konstantin Shishkov
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #ifndef AVCODEC_PSYMODEL_H
23 #define AVCODEC_PSYMODEL_H
27 /** maximum possible number of bands */
28 #define PSY_MAX_BANDS 128
29 /** maximum number of channels */
30 #define PSY_MAX_CHANS 20
32 /* cutoff for VBR is purposely increased, since LP filtering actually
33 * hinders VBR performance rather than the opposite
35 #define AAC_CUTOFF_FROM_BITRATE(bit_rate,channels,sample_rate) (bit_rate ? FFMIN3(FFMIN3( \
36 FFMAX(bit_rate/channels/5, bit_rate/channels*15/32 - 5500), \
37 3000 + bit_rate/channels/4, \
38 12000 + bit_rate/channels/16), \
40 sample_rate / 2): (sample_rate / 2))
41 #define AAC_CUTOFF(s) ( \
42 (s->flags & AV_CODEC_FLAG_QSCALE) \
43 ? s->sample_rate / 2 \
44 : AAC_CUTOFF_FROM_BITRATE(s->bit_rate, s->channels, s->sample_rate) \
48 * single band psychoacoustic information
50 typedef struct FFPsyBand {
54 float spread; /* Energy spread over the band */
58 * single channel psychoacoustic information
60 typedef struct FFPsyChannel {
61 FFPsyBand psy_bands[PSY_MAX_BANDS]; ///< channel bands information
62 float entropy; ///< total PE for this channel
66 * psychoacoustic information for an arbitrary group of channels
68 typedef struct FFPsyChannelGroup {
69 FFPsyChannel *ch[PSY_MAX_CHANS]; ///< pointers to the individual channels in the group
70 uint8_t num_ch; ///< number of channels in this group
71 uint8_t coupling[PSY_MAX_BANDS]; ///< allow coupling for this band in the group
75 * windowing related information
77 typedef struct FFPsyWindowInfo {
78 int window_type[3]; ///< window type (short/long/transitional, etc.) - current, previous and next
79 int window_shape; ///< window shape (sine/KBD/whatever)
80 int num_windows; ///< number of windows in a frame
81 int grouping[8]; ///< window grouping (for e.g. AAC)
82 float clipping[8]; ///< maximum absolute normalized intensity in the given window for clip avoidance
83 int *window_sizes; ///< sequence of window sizes inside one frame (for eg. WMA)
87 * context used by psychoacoustic model
89 typedef struct FFPsyContext {
90 AVCodecContext *avctx; ///< encoder context
91 const struct FFPsyModel *model; ///< encoder-specific model functions
93 FFPsyChannel *ch; ///< single channel information
94 FFPsyChannelGroup *group; ///< channel group information
95 int num_groups; ///< number of channel groups
96 int cutoff; ///< lowpass frequency cutoff for analysis
98 uint8_t **bands; ///< scalefactor band sizes for possible frame sizes
99 int *num_bands; ///< number of scalefactor bands for possible frame sizes
100 int num_lens; ///< number of scalefactor band sets
103 int size; ///< size of the bitresevoir in bits
104 int bits; ///< number of bits used in the bitresevoir
105 int alloc; ///< number of bits allocated by the psy, or -1 if no allocation was done
108 void* model_priv_data; ///< psychoacoustic model implementation private data
112 * codec-specific psychoacoustic model implementation
114 typedef struct FFPsyModel {
116 int (*init) (FFPsyContext *apc);
119 * Suggest window sequence for channel.
121 * @param ctx model context
122 * @param audio samples for the current frame
123 * @param la lookahead samples (NULL when unavailable)
124 * @param channel number of channel element to analyze
125 * @param prev_type previous window type
127 * @return suggested window information in a structure
129 FFPsyWindowInfo (*window)(FFPsyContext *ctx, const float *audio, const float *la, int channel, int prev_type);
132 * Perform psychoacoustic analysis and set band info (threshold, energy) for a group of channels.
134 * @param ctx model context
135 * @param channel channel number of the first channel in the group to perform analysis on
136 * @param coeffs array of pointers to the transformed coefficients
137 * @param wi window information for the channels in the group
139 void (*analyze)(FFPsyContext *ctx, int channel, const float **coeffs, const FFPsyWindowInfo *wi);
141 void (*end) (FFPsyContext *apc);
145 * Initialize psychoacoustic model.
147 * @param ctx model context
148 * @param avctx codec context
149 * @param num_lens number of possible frame lengths
150 * @param bands scalefactor band lengths for all frame lengths
151 * @param num_bands number of scalefactor bands for all frame lengths
152 * @param num_groups number of channel groups
153 * @param group_map array with # of channels in group - 1, for each group
155 * @return zero if successful, a negative value if not
157 int ff_psy_init(FFPsyContext *ctx, AVCodecContext *avctx, int num_lens,
158 const uint8_t **bands, const int *num_bands,
159 int num_groups, const uint8_t *group_map);
162 * Determine what group a channel belongs to.
164 * @param ctx psymodel context
165 * @param channel channel to locate the group for
167 * @return pointer to the FFPsyChannelGroup this channel belongs to
169 FFPsyChannelGroup *ff_psy_find_group(FFPsyContext *ctx, int channel);
172 * Cleanup model context at the end.
174 * @param ctx model context
176 void ff_psy_end(FFPsyContext *ctx);
179 /**************************************************************************
180 * Audio preprocessing stuff. *
181 * This should be moved into some audio filter eventually. *
182 **************************************************************************/
183 struct FFPsyPreprocessContext;
186 * psychoacoustic model audio preprocessing initialization
188 struct FFPsyPreprocessContext *ff_psy_preprocess_init(AVCodecContext *avctx);
191 * Preprocess several channel in audio frame in order to compress it better.
193 * @param ctx preprocessing context
194 * @param audio samples to be filtered (in place)
195 * @param channels number of channel to preprocess
197 void ff_psy_preprocess(struct FFPsyPreprocessContext *ctx, float **audio, int channels);
200 * Cleanup audio preprocessing module.
202 void ff_psy_preprocess_end(struct FFPsyPreprocessContext *ctx);
204 #endif /* AVCODEC_PSYMODEL_H */