LSF (Line Spectral Frequencies) decoding routine.
[ffmpeg.git] / libavcodec / g729dec.c
index e8656bc..628aee0 100644 (file)
 
 #include "avcodec.h"
 #include "libavutil/avutil.h"
 
 #include "avcodec.h"
 #include "libavutil/avutil.h"
-#include "bitstream.h"
+#include "get_bits.h"
+
+#include "g729.h"
+#include "lsp.h"
+#include "celp_math.h"
+#include "acelp_filters.h"
+#include "acelp_pitch_delay.h"
+#include "acelp_vectors.h"
+#include "g729data.h"
 
 /**
  * minimum quantized LSF value (3.2.4)
 
 /**
  * minimum quantized LSF value (3.2.4)
  */
 #define SHARP_MAX                  13017
 
  */
 #define SHARP_MAX                  13017
 
+typedef struct {
+    uint8_t ac_index_bits[2];   ///< adaptive codebook index for second subframe (size in bits)
+    uint8_t parity_bit;         ///< parity bit for pitch delay
+    uint8_t gc_1st_index_bits;  ///< gain codebook (first stage) index (size in bits)
+    uint8_t gc_2nd_index_bits;  ///< gain codebook (second stage) index (size in bits)
+    uint8_t fc_signs_bits;      ///< number of pulses in fixed-codebook vector
+    uint8_t fc_indexes_bits;    ///< size (in bits) of fixed-codebook index entry
+} G729FormatDescription;
+
+typedef struct {
+    /// (2.13) LSP quantizer outputs
+    int16_t  past_quantizer_output_buf[MA_NP + 1][10];
+    int16_t* past_quantizer_outputs[MA_NP + 1];
+
+    int16_t lsfq[10];           ///< (2.13) quantized LSF coefficients from previous frame
+    int16_t lsp_buf[2][10];     ///< (0.15) LSP coefficients (previous and current frames) (3.2.5)
+    int16_t *lsp[2];            ///< pointers to lsp_buf
+}  G729Context;
+
+static const G729FormatDescription format_g729_8k = {
+    .ac_index_bits     = {8,5},
+    .parity_bit        = 1,
+    .gc_1st_index_bits = GC_1ST_IDX_BITS_8K,
+    .gc_2nd_index_bits = GC_2ND_IDX_BITS_8K,
+    .fc_signs_bits     = 4,
+    .fc_indexes_bits   = 13,
+};
+
+static const G729FormatDescription format_g729d_6k4 = {
+    .ac_index_bits     = {8,4},
+    .parity_bit        = 0,
+    .gc_1st_index_bits = GC_1ST_IDX_BITS_6K4,
+    .gc_2nd_index_bits = GC_2ND_IDX_BITS_6K4,
+    .fc_signs_bits     = 2,
+    .fc_indexes_bits   = 9,
+};
+
 /**
  * \brief pseudo random number generator
  */
 /**
  * \brief pseudo random number generator
  */
-static inline uint16_t g729_random(uint16_t value)
+static inline uint16_t g729_prng(uint16_t value)
 {
     return 31821 * value + 13849;
 }
 {
     return 31821 * value + 13849;
 }
@@ -74,45 +119,183 @@ static inline uint16_t g729_random(uint16_t value)
 /**
  * Get parity bit of bit 2..7
  */
 /**
  * Get parity bit of bit 2..7
  */
-static inline int g729_get_parity(uint8_t value)
+static inline int get_parity(uint8_t value)
 {
    return (0x6996966996696996ULL >> (value >> 2)) & 1;
 }
 
 {
    return (0x6996966996696996ULL >> (value >> 2)) & 1;
 }
 
-        ff_acelp_weighted_vector_sum(
-                fc + pitch_delay_int[i],
-                fc + pitch_delay_int[i],
-                fc,
-                1 << 14,
-                av_clip(ctx->gain_pitch, SHARP_MIN, SHARP_MAX),
-                0,
-                14,
-                ctx->subframe_size - pitch_delay_int[i]);
-
-            ctx->gain_pitch  = cb_gain_1st_8k[parm->gc_1st_index[i]][0] +
-                               cb_gain_2nd_8k[parm->gc_2nd_index[i]][0];
-            gain_corr_factor = cb_gain_1st_8k[parm->gc_1st_index[i]][1] +
-                               cb_gain_2nd_8k[parm->gc_2nd_index[i]][1];
-
-        ff_acelp_weighted_vector_sum(
-                ctx->exc + i * ctx->subframe_size,
-                ctx->exc + i * ctx->subframe_size,
-                fc,
-                (!voicing && ctx->frame_erasure) ? 0 : ctx->gain_pitch,
-                ( voicing && ctx->frame_erasure) ? 0 : ctx->gain_code,
-                1<<13,
-                14,
-                ctx->subframe_size);
+static void lsf_decode(int16_t* lsfq, int16_t* past_quantizer_outputs[MA_NP + 1],
+                       int16_t ma_predictor,
+                       int16_t vq_1st, int16_t vq_2nd_low, int16_t vq_2nd_high)
+{
+    int i,j;
+    static const uint8_t min_distance[2]={10, 5}; //(2.13)
+    int16_t* quantizer_output = past_quantizer_outputs[MA_NP];
+
+    for (i = 0; i < 5; i++) {
+        quantizer_output[i]     = cb_lsp_1st[vq_1st][i    ] + cb_lsp_2nd[vq_2nd_low ][i    ];
+        quantizer_output[i + 5] = cb_lsp_1st[vq_1st][i + 5] + cb_lsp_2nd[vq_2nd_high][i + 5];
+    }
+
+    for (j = 0; j < 2; j++) {
+        for (i = 1; i < 10; i++) {
+            int diff = (quantizer_output[i - 1] - quantizer_output[i] + min_distance[j]) >> 1;
+            if (diff > 0) {
+                quantizer_output[i - 1] -= diff;
+                quantizer_output[i    ] += diff;
+            }
+        }
+    }
+
+    for (i = 0; i < 10; i++) {
+        int sum = quantizer_output[i] * cb_ma_predictor_sum[ma_predictor][i];
+        for (j = 0; j < MA_NP; j++)
+            sum += past_quantizer_outputs[j][i] * cb_ma_predictor[ma_predictor][j][i];
+
+        lsfq[i] = sum >> 15;
+    }
+
+    /* Rotate past_quantizer_outputs. */
+    memmove(past_quantizer_outputs + 1, past_quantizer_outputs, MA_NP * sizeof(int16_t*));
+    past_quantizer_outputs[0] = quantizer_output;
+
+    ff_acelp_reorder_lsf(lsfq, LSFQ_DIFF_MIN, LSFQ_MIN, LSFQ_MAX, 10);
+}
+
+static av_cold int decoder_init(AVCodecContext * avctx)
+{
+    G729Context* ctx = avctx->priv_data;
+    int i,k;
+
+    if (avctx->channels != 1) {
+        av_log(avctx, AV_LOG_ERROR, "Only mono sound is supported (requested channels: %d).\n", avctx->channels);
+        return AVERROR_NOFMT;
+    }
+
+    /* Both 8kbit/s and 6.4kbit/s modes uses two subframes per frame. */
+    avctx->frame_size = SUBFRAME_SIZE << 1;
+
+    for (k = 0; k < MA_NP + 1; k++) {
+        ctx->past_quantizer_outputs[k] = ctx->past_quantizer_output_buf[k];
+        for (i = 1; i < 11; i++)
+            ctx->past_quantizer_outputs[k][i - 1] = (18717 * i) >> 3;
+    }
+
+    ctx->lsp[0] = ctx->lsp_buf[0];
+    ctx->lsp[1] = ctx->lsp_buf[1];
+    memcpy(ctx->lsp[0], lsp_init, 10 * sizeof(int16_t));
+
+    return 0;
+}
+
+static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
+                        AVPacket *avpkt)
+{
+    const uint8_t *buf = avpkt->data;
+    int buf_size       = avpkt->size;
+    int16_t *out_frame = data;
+    GetBitContext gb;
+    G729FormatDescription format;
+    int frame_erasure = 0;    ///< frame erasure detected during decoding
+    int bad_pitch = 0;        ///< parity check failed
+    int i;
+    G729Context *ctx = avctx->priv_data;
+    int16_t lp[2][11];           // (3.12)
+    uint8_t ma_predictor;     ///< switched MA predictor of LSP quantizer
+    uint8_t quantizer_1st;    ///< first stage vector of quantizer
+    uint8_t quantizer_2nd_lo; ///< second stage lower vector of quantizer (size in bits)
+    uint8_t quantizer_2nd_hi; ///< second stage higher vector of quantizer (size in bits)
+
+    if (*data_size < SUBFRAME_SIZE << 2) {
+        av_log(avctx, AV_LOG_ERROR, "Error processing packet: output buffer too small\n");
+        return AVERROR(EIO);
+    }
+
+    if (buf_size == 10) {
+        format = format_g729_8k;
+        av_log(avctx, AV_LOG_DEBUG, "Packet type: %s\n", "G.729 @ 8kbit/s");
+    } else if (buf_size == 8) {
+        format = format_g729d_6k4;
+        av_log(avctx, AV_LOG_DEBUG, "Packet type: %s\n", "G.729D @ 6.4kbit/s");
+    } else {
+        av_log(avctx, AV_LOG_ERROR, "Packet size %d is unknown.\n", buf_size);
+        return (AVERROR_NOFMT);
+    }
+
+    for (i=0; i < buf_size; i++)
+        frame_erasure |= buf[i];
+    frame_erasure = !frame_erasure;
+
+    init_get_bits(&gb, buf, buf_size);
+
+    ma_predictor     = get_bits(&gb, 1);
+    quantizer_1st    = get_bits(&gb, VQ_1ST_BITS);
+    quantizer_2nd_lo = get_bits(&gb, VQ_2ND_BITS);
+    quantizer_2nd_hi = get_bits(&gb, VQ_2ND_BITS);
+
+    lsf_decode(ctx->lsfq, ctx->past_quantizer_outputs,
+               ma_predictor,
+               quantizer_1st, quantizer_2nd_lo, quantizer_2nd_hi);
+
+    ff_acelp_lsf2lsp(ctx->lsp[1], ctx->lsfq, 10);
+
+    ff_acelp_lp_decode(&lp[0][0], &lp[1][0], ctx->lsp[1], ctx->lsp[0], 10);
+
+    FFSWAP(int16_t*, ctx->lsp[1], ctx->lsp[0]);
+
+    for (i = 0; i < 2; i++) {
+        uint8_t ac_index;      ///< adaptive codebook index
+        uint8_t pulses_signs;  ///< fixed-codebook vector pulse signs
+        int fc_indexes;        ///< fixed-codebook indexes
+        uint8_t gc_1st_index;  ///< gain codebook (first stage) index
+        uint8_t gc_2nd_index;  ///< gain codebook (second stage) index
+
+        ac_index      = get_bits(&gb, format.ac_index_bits[i]);
+        if(!i && format.parity_bit)
+            bad_pitch = get_parity(ac_index) == get_bits1(&gb);
+        fc_indexes    = get_bits(&gb, format.fc_indexes_bits);
+        pulses_signs  = get_bits(&gb, format.fc_signs_bits);
+        gc_1st_index  = get_bits(&gb, format.gc_1st_index_bits);
+        gc_2nd_index  = get_bits(&gb, format.gc_2nd_index_bits);
+
+        ff_acelp_weighted_vector_sum(fc + pitch_delay_int[i],
+                                     fc + pitch_delay_int[i],
+                                     fc, 1 << 14,
+                                     av_clip(ctx->gain_pitch, SHARP_MIN, SHARP_MAX),
+                                     0, 14,
+                                     SUBFRAME_SIZE - pitch_delay_int[i]);
+
+        if (frame_erasure) {
+            ctx->gain_pitch = (29491 * ctx->gain_pitch) >> 15; // 0.90 (0.15)
+            ctx->gain_code  = ( 2007 * ctx->gain_code ) >> 11; // 0.98 (0.11)
+
+            gain_corr_factor = 0;
+        } else {
+            ctx->gain_pitch  = cb_gain_1st_8k[gc_1st_index][0] +
+                               cb_gain_2nd_8k[gc_2nd_index][0];
+            gain_corr_factor = cb_gain_1st_8k[gc_1st_index][1] +
+                               cb_gain_2nd_8k[gc_2nd_index][1];
+
+        ff_acelp_weighted_vector_sum(ctx->exc + i * SUBFRAME_SIZE,
+                                     ctx->exc + i * SUBFRAME_SIZE, fc,
+                                     (!voicing && frame_erasure) ? 0 : ctx->gain_pitch,
+                                     ( voicing && frame_erasure) ? 0 : ctx->gain_code,
+                                     1 << 13, 14, SUBFRAME_SIZE);
+    }
+
+    *data_size = SUBFRAME_SIZE << 2;
+    return buf_size;
+}
 
 AVCodec g729_decoder =
 {
     "g729",
     CODEC_TYPE_AUDIO,
     CODEC_ID_G729,
 
 AVCodec g729_decoder =
 {
     "g729",
     CODEC_TYPE_AUDIO,
     CODEC_ID_G729,
-    sizeof(G729_Context),
-    ff_g729_decoder_init,
+    sizeof(G729Context),
+    decoder_init,
     NULL,
     NULL,
     NULL,
     NULL,
-    ff_g729_decode_frame,
+    decode_frame,
     .long_name = NULL_IF_CONFIG_SMALL("G.729"),
 };
     .long_name = NULL_IF_CONFIG_SMALL("G.729"),
 };