Misc SILK fixes:
[opus.git] / silk / structs.h
1 /***********************************************************************
2 Copyright (c) 2006-2011, Skype Limited. All rights reserved.
3 Redistribution and use in source and binary forms, with or without
4 modification, (subject to the limitations in the disclaimer below)
5 are permitted provided that the following conditions are met:
6 - Redistributions of source code must retain the above copyright notice,
7 this list of conditions and the following disclaimer.
8 - Redistributions in binary form must reproduce the above copyright
9 notice, this list of conditions and the following disclaimer in the
10 documentation and/or other materials provided with the distribution.
11 - Neither the name of Skype Limited, nor the names of specific
12 contributors, may be used to endorse or promote products derived from
13 this software without specific prior written permission.
14 NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY'S PATENT RIGHTS ARE GRANTED
15 BY THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
16 CONTRIBUTORS ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
17 BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
18 FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
19 COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
20 INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
22 USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
23 ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
25 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 ***********************************************************************/
27
28 #ifndef SILK_STRUCTS_H
29 #define SILK_STRUCTS_H
30
31 #include "typedef.h"
32 #include "SigProc_FIX.h"
33 #include "define.h"
34 #include "entenc.h"
35 #include "entdec.h"
36
37 #ifdef __cplusplus
38 extern "C"
39 {
40 #endif
41
42 /************************************/
43 /* Noise shaping quantization state */
44 /************************************/
45 typedef struct {
46     opus_int16   xq[           2 * MAX_FRAME_LENGTH ]; /* Buffer for quantized output signal */
47     opus_int32   sLTP_shp_Q10[ 2 * MAX_FRAME_LENGTH ];
48     opus_int32   sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ];
49     opus_int32   sAR2_Q14[ MAX_SHAPE_LPC_ORDER ];
50     opus_int32   sLF_AR_shp_Q12;
51     opus_int     lagPrev;
52     opus_int     sLTP_buf_idx;
53     opus_int     sLTP_shp_buf_idx;
54     opus_int32   rand_seed;
55     opus_int32   prev_inv_gain_Q16;
56     opus_int     rewhite_flag;
57 } silk_nsq_state;
58
59 /********************************/
60 /* VAD state                    */
61 /********************************/
62 typedef struct {
63     opus_int32   AnaState[ 2 ];                  /* Analysis filterbank state: 0-8 kHz                       */
64     opus_int32   AnaState1[ 2 ];                 /* Analysis filterbank state: 0-4 kHz                       */
65     opus_int32   AnaState2[ 2 ];                 /* Analysis filterbank state: 0-2 kHz                       */
66     opus_int32   XnrgSubfr[ VAD_N_BANDS ];       /* Subframe energies                                        */
67     opus_int32   NrgRatioSmth_Q8[ VAD_N_BANDS ]; /* Smoothed energy level in each band                       */
68     opus_int16   HPstate;                        /* State of differentiator in the lowest band               */
69     opus_int32   NL[ VAD_N_BANDS ];              /* Noise energy level in each band                          */
70     opus_int32   inv_NL[ VAD_N_BANDS ];          /* Inverse noise energy level in each band                  */
71     opus_int32   NoiseLevelBias[ VAD_N_BANDS ];  /* Noise level estimator bias/offset                        */
72     opus_int32   counter;                        /* Frame counter used in the initial phase                  */
73 } silk_VAD_state;
74
75 /* Variable cut-off low-pass filter state */
76 typedef struct {
77     opus_int32                   In_LP_State[ 2 ];           /* Low pass filter state */
78     opus_int32                   transition_frame_no;        /* Counter which is mapped to a cut-off frequency */
79     opus_int                     mode;                       /* Operating mode, <0: switch down, >0: switch up; 0: do nothing */
80 } silk_LP_state;
81
82 /* Structure containing NLSF codebook */
83 typedef struct {
84     const opus_int16             nVectors;
85     const opus_int16             order;
86     const opus_int16             quantStepSize_Q16;
87     const opus_int16             invQuantStepSize_Q6;
88     const opus_uint8             *CB1_NLSF_Q8;
89     const opus_uint8             *CB1_iCDF;
90     const opus_uint8             *pred_Q8;
91     const opus_uint8             *ec_sel;
92     const opus_uint8             *ec_iCDF;
93     const opus_uint8             *ec_Rates_Q5;
94     const opus_int16             *deltaMin_Q15;
95 } silk_NLSF_CB_struct;
96
97 typedef struct {
98     opus_int16                   pred_prev_Q13[ 2 ];
99     opus_int16                   sMid[ 2 ];
100     opus_int16                   sSide[ 2 ];
101     opus_int32                   mid_side_amp_Q0[ 4 ];
102     opus_int16                   smth_width_Q14;
103     opus_int16                   width_prev_Q14;
104     opus_int8                    predIx[ MAX_FRAMES_PER_PACKET ][ 2 ][ 3 ];
105     opus_int8                    mid_only_flags[ MAX_FRAMES_PER_PACKET ];
106 } stereo_enc_state;
107
108 typedef struct {
109     opus_int16                   pred_prev_Q13[ 2 ];
110     opus_int16                   sMid[ 2 ];
111     opus_int16                   sSide[ 2 ];
112 } stereo_dec_state;
113
114 typedef struct {
115     opus_int8                    GainsIndices[ MAX_NB_SUBFR ];
116     opus_int8                    LTPIndex[ MAX_NB_SUBFR ];
117     opus_int8                    NLSFIndices[ MAX_LPC_ORDER + 1 ];
118     opus_int16                   lagIndex;
119     opus_int8                    contourIndex;
120     opus_int8                    signalType;
121     opus_int8                    quantOffsetType;
122     opus_int8                    NLSFInterpCoef_Q2;
123     opus_int8                    PERIndex;
124     opus_int8                    LTP_scaleIndex;
125     opus_int8                    Seed;
126 } SideInfoIndices;
127
128 /********************************/
129 /* Encoder state                */
130 /********************************/
131 typedef struct {
132     opus_int32                       In_HP_State[ 2 ];               /* High pass filter state                                               */
133     opus_int32                       variable_HP_smth1_Q15;          /* State of first smoother                                              */
134     opus_int32                       variable_HP_smth2_Q15;          /* State of second smoother                                             */
135     silk_LP_state                   sLP;                            /* Low pass filter state                                                */
136     silk_VAD_state                  sVAD;                           /* Voice activity detector state                                        */
137     silk_nsq_state                  sNSQ;                           /* Noise Shape Quantizer State                                          */
138     opus_int16                       prev_NLSFq_Q15[ MAX_LPC_ORDER ];/* Previously quantized NLSF vector                                     */
139     opus_int                         speech_activity_Q8;             /* Speech activity                                                      */
140     opus_int                         allow_bandwidth_switch;         /* Flag indicating that switching of internal bandwidth is allowed      */
141     opus_int8                        LBRRprevLastGainIndex;
142     opus_int8                        prevSignalType;
143     opus_int                         prevLag;
144     opus_int                         pitch_LPC_win_length;
145     opus_int                         max_pitch_lag;                  /* Highest possible pitch lag (samples)                                 */
146     opus_int32                       API_fs_Hz;                      /* API sampling frequency (Hz)                                          */
147     opus_int32                       prev_API_fs_Hz;                 /* Previous API sampling frequency (Hz)                                 */
148     opus_int                         maxInternal_fs_Hz;              /* Maximum internal sampling frequency (Hz)                             */
149     opus_int                         minInternal_fs_Hz;              /* Minimum internal sampling frequency (Hz)                             */
150     opus_int                         desiredInternal_fs_Hz;          /* Soft request for internal sampling frequency (Hz)                    */
151     opus_int                         fs_kHz;                         /* Internal sampling frequency (kHz)                                    */
152     opus_int                         delay;                          /* Number of samples of delay to apply */
153     opus_int                         nb_subfr;                       /* Number of 5 ms subframes in a frame                                  */
154     opus_int                         frame_length;                   /* Frame length (samples)                                               */
155     opus_int                         subfr_length;                   /* Subframe length (samples)                                            */
156     opus_int                         ltp_mem_length;                 /* Length of LTP memory                                                 */
157     opus_int                         la_pitch;                       /* Look-ahead for pitch analysis (samples)                              */
158     opus_int                         la_shape;                       /* Look-ahead for noise shape analysis (samples)                        */
159     opus_int                         shapeWinLength;                 /* Window length for noise shape analysis (samples)                     */
160     opus_int32                       TargetRate_bps;                 /* Target bitrate (bps)                                                 */
161     opus_int                         PacketSize_ms;                  /* Number of milliseconds to put in each packet                         */
162     opus_int                         PacketLoss_perc;                /* Packet loss rate measured by farend                                  */
163     opus_int32                       frameCounter;
164     opus_int                         Complexity;                     /* Complexity setting                                                   */
165     opus_int                         nStatesDelayedDecision;         /* Number of states in delayed decision quantization                    */
166     opus_int                         useInterpolatedNLSFs;           /* Flag for using NLSF interpolation                                    */
167     opus_int                         shapingLPCOrder;                /* Filter order for noise shaping filters                               */
168     opus_int                         predictLPCOrder;                /* Filter order for prediction filters                                  */
169     opus_int                         pitchEstimationComplexity;      /* Complexity level for pitch estimator                                 */
170     opus_int                         pitchEstimationLPCOrder;        /* Whitening filter order for pitch estimator                           */
171     opus_int32                       pitchEstimationThreshold_Q16;   /* Threshold for pitch estimator                                        */
172     opus_int                         LTPQuantLowComplexity;          /* Flag for low complexity LTP quantization                             */
173     opus_int                         mu_LTP_Q9;                      /* Rate-distortion tradeoff in LTP quantization                         */
174     opus_int                         NLSF_MSVQ_Survivors;            /* Number of survivors in NLSF MSVQ                                     */
175     opus_int                         first_frame_after_reset;        /* Flag for deactivating NLSF interp. and fluc. reduction after resets  */
176     opus_int                         controlled_since_last_payload;  /* Flag for ensuring codec_control only runs once per packet            */
177     opus_int                         warping_Q16;                    /* Warping parameter for warped noise shaping                           */
178     opus_int                         useCBR;                         /* Flag to enable constant bitrate                                      */
179     opus_int                         prefillFlag;                    /* Flag to indicate that only buffers are prefilled, no coding          */
180     const opus_uint8                 *pitch_lag_low_bits_iCDF;       /* Pointer to iCDF table for low bits of pitch lag index                */
181     const opus_uint8                 *pitch_contour_iCDF;            /* Pointer to iCDF table for pitch contour index                        */
182     const silk_NLSF_CB_struct       *psNLSF_CB;                     /* Pointer to NLSF codebook                                             */
183     opus_int                         input_quality_bands_Q15[ VAD_N_BANDS ];
184     opus_int                         input_tilt_Q15;
185     opus_int                         SNR_dB_Q7;                      /* Quality setting                                                      */
186
187     opus_int8                        VAD_flags[ MAX_FRAMES_PER_PACKET ];
188     opus_int8                        LBRR_flag;
189     opus_int                         LBRR_flags[ MAX_FRAMES_PER_PACKET ];
190
191     SideInfoIndices                 indices;
192     opus_int8                        pulses[ MAX_FRAME_LENGTH ];
193
194     /* Input/output buffering */
195     opus_int16                       inputBuf[ MAX_FRAME_LENGTH + 2 ]; /* Buffer containing input signal                                   */
196     opus_int16                       delayBuf[MAX_ENCODER_DELAY];
197     opus_int                         inputBufIx;
198     opus_int                         nFramesPerPacket;
199     opus_int                         nFramesEncoded;                 /* Number of frames analyzed in current packet                          */
200
201     opus_int                         nChannelsAPI;
202     opus_int                         nChannelsInternal;
203     opus_int                         channelNb;
204
205     /* Parameters For LTP scaling Control */
206     opus_int                         frames_since_onset;
207
208     /* Specifically for entropy coding */
209     opus_int                         ec_prevSignalType;
210     opus_int16                       ec_prevLagIndex;
211
212     silk_resampler_state_struct resampler_state;
213
214     /* DTX */
215     opus_int                         useDTX;                         /* Flag to enable DTX                                                   */
216     opus_int                         inDTX;                          /* Flag to signal DTX period                                            */
217     opus_int                         noSpeechCounter;                /* Counts concecutive nonactive frames, used by DTX                     */
218
219     /* Inband Low Bitrate Redundancy (LBRR) data */
220     opus_int                         useInBandFEC;                   /* Saves the API setting for query                                      */
221     opus_int                         LBRR_enabled;                   /* Depends on useInBandFRC, bitrate and packet loss rate                */
222     opus_int                         LBRR_GainIncreases;             /* Gains increment for coding LBRR frames                               */
223     SideInfoIndices                 indices_LBRR[ MAX_FRAMES_PER_PACKET ];
224     opus_int8                        pulses_LBRR[ MAX_FRAMES_PER_PACKET ][ MAX_FRAME_LENGTH ];
225 } silk_encoder_state;
226
227
228 /* Struct for Packet Loss Concealment */
229 typedef struct {
230     opus_int32   pitchL_Q8;                      /* Pitch lag to use for voiced concealment                  */
231     opus_int16   LTPCoef_Q14[ LTP_ORDER ];       /* LTP coeficients to use for voiced concealment            */
232     opus_int16   prevLPC_Q12[ MAX_LPC_ORDER ];
233     opus_int     last_frame_lost;                /* Was previous frame lost                                  */
234     opus_int32   rand_seed;                      /* Seed for unvoiced signal generation                      */
235     opus_int16   randScale_Q14;                  /* Scaling of unvoiced random signal                        */
236     opus_int32   conc_energy;
237     opus_int     conc_energy_shift;
238     opus_int16   prevLTP_scale_Q14;
239     opus_int32   prevGain_Q16[ MAX_NB_SUBFR ];
240     opus_int     fs_kHz;
241 } silk_PLC_struct;
242
243 /* Struct for CNG */
244 typedef struct {
245     opus_int32   CNG_exc_buf_Q10[ MAX_FRAME_LENGTH ];
246     opus_int16   CNG_smth_NLSF_Q15[ MAX_LPC_ORDER ];
247     opus_int32   CNG_synth_state[ MAX_LPC_ORDER ];
248     opus_int32   CNG_smth_Gain_Q16;
249     opus_int32   rand_seed;
250     opus_int     fs_kHz;
251 } silk_CNG_struct;
252
253 /********************************/
254 /* Decoder state                */
255 /********************************/
256 typedef struct {
257     opus_int32       prev_inv_gain_Q16;
258     opus_int32       exc_Q10[ MAX_FRAME_LENGTH ];
259     opus_int32       sLPC_Q14_buf[ MAX_LPC_ORDER ];
260     opus_int16       outBuf[ MAX_FRAME_LENGTH + 2 * MAX_SUB_FRAME_LENGTH ];  /* Buffer for output signal                                 */
261     opus_int16       delayBuf[ MAX_DECODER_DELAY ];              /* Buffer for delaying the SILK output prior to resampling              */
262     opus_int         delay;                                      /* How much decoder delay to add                                        */
263     opus_int         lagPrev;                                    /* Previous Lag                                                         */
264     opus_int8        LastGainIndex;                              /* Previous gain index                                                  */
265     opus_int         fs_kHz;                                     /* Sampling frequency in kHz                                            */
266     opus_int32       prev_API_sampleRate;                        /* Previous API sample frequency (Hz)                                   */
267     opus_int         nb_subfr;                                   /* Number of 5 ms subframes in a frame                                  */
268     opus_int         frame_length;                               /* Frame length (samples)                                               */
269     opus_int         subfr_length;                               /* Subframe length (samples)                                            */
270     opus_int         ltp_mem_length;                             /* Length of LTP memory                                                 */
271     opus_int         LPC_order;                                  /* LPC order                                                            */
272     opus_int16       prevNLSF_Q15[ MAX_LPC_ORDER ];              /* Used to interpolate LSFs                                             */
273     opus_int         first_frame_after_reset;                    /* Flag for deactivating NLSF interp. and fluc. reduction after resets  */
274     const opus_uint8 *pitch_lag_low_bits_iCDF;                   /* Pointer to iCDF table for low bits of pitch lag index                */
275     const opus_uint8 *pitch_contour_iCDF;                        /* Pointer to iCDF table for pitch contour index                        */
276
277     /* For buffering payload in case of more frames per packet */
278     opus_int         nFramesDecoded;
279     opus_int         nFramesPerPacket;
280
281     /* Specifically for entropy coding */
282     opus_int         ec_prevSignalType;
283     opus_int16       ec_prevLagIndex;
284
285     opus_int         VAD_flags[ MAX_FRAMES_PER_PACKET ];
286     opus_int         LBRR_flag;
287     opus_int         LBRR_flags[ MAX_FRAMES_PER_PACKET ];
288
289     silk_resampler_state_struct resampler_state;
290
291     const silk_NLSF_CB_struct   *psNLSF_CB;                 /* Pointer to NLSF codebook                                             */
292
293     /* Quantization indices */
294     SideInfoIndices indices;
295
296     /* CNG state */
297     silk_CNG_struct sCNG;
298
299     /* Stuff used for PLC */
300     opus_int         lossCnt;
301     opus_int         prevSignalType;
302
303     silk_PLC_struct sPLC;
304
305 } silk_decoder_state;
306
307 /************************/
308 /* Decoder control      */
309 /************************/
310 typedef struct {
311     /* prediction and coding parameters */
312     opus_int             pitchL[ MAX_NB_SUBFR ];
313     opus_int32           Gains_Q16[ MAX_NB_SUBFR ];
314     /* holds interpolated and final coefficients, 4-byte aligned */
315     silk_DWORD_ALIGN opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ];
316     opus_int16           LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ];
317     opus_int             LTP_scale_Q14;
318 } silk_decoder_control;
319
320
321 #ifdef __cplusplus
322 }
323 #endif
324
325 #endif