diff options
Diffstat (limited to 'pjmedia/src/pjmedia-codec/speex/nb_celp.h')
| -rw-r--r-- | pjmedia/src/pjmedia-codec/speex/nb_celp.h | 159 |
1 files changed, 74 insertions, 85 deletions
diff --git a/pjmedia/src/pjmedia-codec/speex/nb_celp.h b/pjmedia/src/pjmedia-codec/speex/nb_celp.h index c9fb2b3d..92028cb5 100644 --- a/pjmedia/src/pjmedia-codec/speex/nb_celp.h +++ b/pjmedia/src/pjmedia-codec/speex/nb_celp.h @@ -48,20 +48,20 @@ /**Structure representing the full state of the narrowband encoder*/ typedef struct EncState { - const SpeexMode *mode; /**< Mode corresponding to the state */ - int first; /**< Is this the first frame? */ - int frameSize; /**< Size of frames */ - int subframeSize; /**< Size of sub-frames */ - int nbSubframes; /**< Number of sub-frames */ - int windowSize; /**< Analysis (LPC) window length */ - int lpcSize; /**< LPC order */ - int min_pitch; /**< Minimum pitch value allowed */ - int max_pitch; /**< Maximum pitch value allowed */ - - int safe_pitch; /**< Don't use too large values for pitch (in case we lose a packet) */ - int bounded_pitch; /**< Next frame should not rely on previous frames for pitch */ - int ol_pitch; /**< Open-loop pitch */ - int ol_voiced; /**< Open-loop voiced/non-voiced decision */ + const SpeexMode *mode; /**< Mode corresponding to the state */ + int first; /**< Is this the first frame? */ + int frameSize; /**< Size of frames */ + int subframeSize; /**< Size of sub-frames */ + int nbSubframes; /**< Number of sub-frames */ + int windowSize; /**< Analysis (LPC) window length */ + int lpcSize; /**< LPC order */ + int min_pitch; /**< Minimum pitch value allowed */ + int max_pitch; /**< Maximum pitch value allowed */ + + spx_word32_t cumul_gain; /**< Product of previously used pitch gains (Q10) */ + int bounded_pitch; /**< Next frame should not rely on previous frames for pitch */ + int ol_pitch; /**< Open-loop pitch */ + int ol_voiced; /**< Open-loop voiced/non-voiced decision */ int *pitch; #ifdef EPIC_48K @@ -70,111 +70,100 @@ typedef struct EncState { #ifdef VORBIS_PSYCHO VorbisPsy *psy; + float *psy_window; float *curve; float *old_curve; #endif spx_word16_t gamma1; /**< Perceptual filter: A(z/gamma1) */ spx_word16_t gamma2; /**< Perceptual filter: A(z/gamma2) */ - float lag_factor; /**< Lag windowing Gaussian width */ + float lag_factor; /**< Lag windowing Gaussian width */ spx_word16_t lpc_floor; /**< Noise floor multiplier for A[0] in LPC analysis*/ - char *stack; /**< Pseudo-stack allocation for temporary memory */ - spx_sig_t *inBuf; /**< Input buffer (original signal) */ - spx_sig_t *frame; /**< Start of original frame */ - spx_sig_t *excBuf; /**< Excitation buffer */ - spx_sig_t *exc; /**< Start of excitation frame */ - spx_sig_t *swBuf; /**< Weighted signal buffer */ - spx_sig_t *sw; /**< Start of weighted signal frame */ - spx_sig_t *innov; /**< Innovation for the frame */ - spx_word16_t *window; /**< Temporary (Hanning) window */ - spx_word16_t *autocorr; /**< auto-correlation */ + char *stack; /**< Pseudo-stack allocation for temporary memory */ + spx_word16_t *winBuf; /**< Input buffer (original signal) */ + spx_word16_t *excBuf; /**< Excitation buffer */ + spx_word16_t *exc; /**< Start of excitation frame */ + spx_word16_t *swBuf; /**< Weighted signal buffer */ + spx_word16_t *sw; /**< Start of weighted signal frame */ + const spx_word16_t *window; /**< Temporary (Hanning) window */ spx_word16_t *lagWindow; /**< Window applied to auto-correlation */ - spx_coef_t *lpc; /**< LPCs for current frame */ - spx_lsp_t *lsp; /**< LSPs for current frame */ - spx_lsp_t *qlsp; /**< Quantized LSPs for current frame */ - spx_lsp_t *old_lsp; /**< LSPs for previous frame */ - spx_lsp_t *old_qlsp; /**< Quantized LSPs for previous frame */ - spx_lsp_t *interp_lsp; /**< Interpolated LSPs */ - spx_lsp_t *interp_qlsp; /**< Interpolated quantized LSPs */ - spx_coef_t *interp_lpc; /**< Interpolated LPCs */ - spx_coef_t *interp_qlpc; /**< Interpolated quantized LPCs */ - spx_coef_t *bw_lpc1; /**< LPCs after bandwidth expansion by gamma1 for perceptual weighting*/ - spx_coef_t *bw_lpc2; /**< LPCs after bandwidth expansion by gamma2 for perceptual weighting*/ - spx_mem_t *mem_sp; /**< Filter memory for signal synthesis */ - spx_mem_t *mem_sw; /**< Filter memory for perceptually-weighted signal */ - spx_mem_t *mem_sw_whole; /**< Filter memory for perceptually-weighted signal (whole frame)*/ - spx_mem_t *mem_exc; /**< Filter memory for excitation (whole frame) */ + spx_lsp_t *old_lsp; /**< LSPs for previous frame */ + spx_lsp_t *old_qlsp; /**< Quantized LSPs for previous frame */ + spx_mem_t *mem_sp; /**< Filter memory for signal synthesis */ + spx_mem_t *mem_sw; /**< Filter memory for perceptually-weighted signal */ + spx_mem_t *mem_sw_whole; /**< Filter memory for perceptually-weighted signal (whole frame)*/ + spx_mem_t *mem_exc; /**< Filter memory for excitation (whole frame) */ + spx_mem_t *mem_exc2; /**< Filter memory for excitation (whole frame) */ spx_word32_t *pi_gain; /**< Gain of LPC filter at theta=pi (fe/2) */ - - VBRState *vbr; /**< State of the VBR data */ - float vbr_quality; /**< Quality setting for VBR encoding */ - float relative_quality; /**< Relative quality that will be needed by VBR */ - int vbr_enabled; /**< 1 for enabling VBR, 0 otherwise */ - int vad_enabled; /**< 1 for enabling VAD, 0 otherwise */ - int dtx_enabled; /**< 1 for enabling DTX, 0 otherwise */ - int dtx_count; /**< Number of consecutive DTX frames */ - int abr_enabled; /**< ABR setting (in bps), 0 if off */ + spx_sig_t *innov_save; /**< If non-NULL, innovation is copied here */ + + VBRState *vbr; /**< State of the VBR data */ + float vbr_quality; /**< Quality setting for VBR encoding */ + float relative_quality; /**< Relative quality that will be needed by VBR */ + int vbr_enabled; /**< 1 for enabling VBR, 0 otherwise */ + spx_int32_t vbr_max; /**< Max bit-rate allowed in VBR mode */ + int vad_enabled; /**< 1 for enabling VAD, 0 otherwise */ + int dtx_enabled; /**< 1 for enabling DTX, 0 otherwise */ + int dtx_count; /**< Number of consecutive DTX frames */ + spx_int32_t abr_enabled; /**< ABR setting (in bps), 0 if off */ float abr_drift; float abr_drift2; float abr_count; - int complexity; /**< Complexity setting (0-10 from least complex to most complex) */ - int sampling_rate; + int complexity; /**< Complexity setting (0-10 from least complex to most complex) */ + spx_int32_t sampling_rate; int plc_tuning; int encode_submode; const SpeexSubmode * const *submodes; /**< Sub-mode data */ - int submodeID; /**< Activated sub-mode */ - int submodeSelect; /**< Mode chosen by the user (may differ from submodeID if VAD is on) */ + int submodeID; /**< Activated sub-mode */ + int submodeSelect; /**< Mode chosen by the user (may differ from submodeID if VAD is on) */ } EncState; /**Structure representing the full state of the narrowband decoder*/ typedef struct DecState { const SpeexMode *mode; /**< Mode corresponding to the state */ - int first; /**< Is this the first frame? */ - int count_lost; /**< Was the last frame lost? */ - int frameSize; /**< Size of frames */ - int subframeSize; /**< Size of sub-frames */ - int nbSubframes; /**< Number of sub-frames */ - int lpcSize; /**< LPC order */ - int min_pitch; /**< Minimum pitch value allowed */ - int max_pitch; /**< Maximum pitch value allowed */ - int sampling_rate; + int first; /**< Is this the first frame? */ + int count_lost; /**< Was the last frame lost? */ + int frameSize; /**< Size of frames */ + int subframeSize; /**< Size of sub-frames */ + int nbSubframes; /**< Number of sub-frames */ + int lpcSize; /**< LPC order */ + int min_pitch; /**< Minimum pitch value allowed */ + int max_pitch; /**< Maximum pitch value allowed */ + spx_int32_t sampling_rate; #ifdef EPIC_48K int lbr_48k; #endif - spx_word16_t last_ol_gain; /**< Open-loop gain for previous frame */ - - char *stack; /**< Pseudo-stack allocation for temporary memory */ - spx_sig_t *inBuf; /**< Input buffer (original signal) */ - spx_sig_t *frame; /**< Start of original frame */ - spx_sig_t *excBuf; /**< Excitation buffer */ - spx_sig_t *exc; /**< Start of excitation frame */ - spx_sig_t *innov; /**< Innovation for the frame */ - spx_lsp_t *qlsp; /**< Quantized LSPs for current frame */ - spx_lsp_t *old_qlsp; /**< Quantized LSPs for previous frame */ - spx_lsp_t *interp_qlsp; /**< Interpolated quantized LSPs */ - spx_coef_t *interp_qlpc; /**< Interpolated quantized LPCs */ - spx_mem_t *mem_sp; /**< Filter memory for synthesis signal */ - spx_word32_t *pi_gain; /**< Gain of LPC filter at theta=pi (fe/2) */ - int last_pitch; /**< Pitch of last correctly decoded frame */ + spx_word16_t last_ol_gain; /**< Open-loop gain for previous frame */ + + char *stack; /**< Pseudo-stack allocation for temporary memory */ + spx_word16_t *excBuf; /**< Excitation buffer */ + spx_word16_t *exc; /**< Start of excitation frame */ + spx_lsp_t *old_qlsp; /**< Quantized LSPs for previous frame */ + spx_coef_t *interp_qlpc; /**< Interpolated quantized LPCs */ + spx_mem_t *mem_sp; /**< Filter memory for synthesis signal */ + spx_word32_t *pi_gain; /**< Gain of LPC filter at theta=pi (fe/2) */ + spx_sig_t *innov_save; /** If non-NULL, innovation is copied here */ + + /* This is used in packet loss concealment */ + int last_pitch; /**< Pitch of last correctly decoded frame */ spx_word16_t last_pitch_gain; /**< Pitch gain of last correctly decoded frame */ - spx_word16_t pitch_gain_buf[3]; /**< Pitch gain of last decoded frames */ - int pitch_gain_buf_idx; /**< Tail of the buffer */ - spx_int32_t seed; /** Seed used for random number generation */ + spx_word16_t pitch_gain_buf[3]; /**< Pitch gain of last decoded frames */ + int pitch_gain_buf_idx; /**< Tail of the buffer */ + spx_int32_t seed; /** Seed used for random number generation */ int encode_submode; const SpeexSubmode * const *submodes; /**< Sub-mode data */ - int submodeID; /**< Activated sub-mode */ - int lpc_enh_enabled; /**< 1 when LPC enhancer is on, 0 otherwise */ - CombFilterMem *comb_mem; + int submodeID; /**< Activated sub-mode */ + int lpc_enh_enabled; /**< 1 when LPC enhancer is on, 0 otherwise */ SpeexCallback speex_callbacks[SPEEX_MAX_CALLBACKS]; SpeexCallback user_callback; /*Vocoder data*/ - float voc_m1; - float voc_m2; + spx_word16_t voc_m1; + spx_word32_t voc_m2; float voc_mean; int voc_offset; |