| 1 | /* |
| 2 | * RealAudio 2.0 (28.8K) |
| 3 | * Copyright (c) 2003 The FFmpeg Project |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 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. |
| 11 | * |
| 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. |
| 16 | * |
| 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 |
| 20 | */ |
| 21 | |
| 22 | #include "libavutil/channel_layout.h" |
| 23 | #include "libavutil/float_dsp.h" |
| 24 | #include "libavutil/internal.h" |
| 25 | #include "avcodec.h" |
| 26 | #include "internal.h" |
| 27 | #define BITSTREAM_READER_LE |
| 28 | #include "get_bits.h" |
| 29 | #include "ra288.h" |
| 30 | #include "lpc.h" |
| 31 | #include "celp_filters.h" |
| 32 | |
| 33 | #define MAX_BACKWARD_FILTER_ORDER 36 |
| 34 | #define MAX_BACKWARD_FILTER_LEN 40 |
| 35 | #define MAX_BACKWARD_FILTER_NONREC 35 |
| 36 | |
| 37 | #define RA288_BLOCK_SIZE 5 |
| 38 | #define RA288_BLOCKS_PER_FRAME 32 |
| 39 | |
| 40 | typedef struct { |
| 41 | AVFloatDSPContext fdsp; |
| 42 | DECLARE_ALIGNED(32, float, sp_lpc)[FFALIGN(36, 16)]; ///< LPC coefficients for speech data (spec: A) |
| 43 | DECLARE_ALIGNED(32, float, gain_lpc)[FFALIGN(10, 16)]; ///< LPC coefficients for gain (spec: GB) |
| 44 | |
| 45 | /** speech data history (spec: SB). |
| 46 | * Its first 70 coefficients are updated only at backward filtering. |
| 47 | */ |
| 48 | float sp_hist[111]; |
| 49 | |
| 50 | /// speech part of the gain autocorrelation (spec: REXP) |
| 51 | float sp_rec[37]; |
| 52 | |
| 53 | /** log-gain history (spec: SBLG). |
| 54 | * Its first 28 coefficients are updated only at backward filtering. |
| 55 | */ |
| 56 | float gain_hist[38]; |
| 57 | |
| 58 | /// recursive part of the gain autocorrelation (spec: REXPLG) |
| 59 | float gain_rec[11]; |
| 60 | } RA288Context; |
| 61 | |
| 62 | static av_cold int ra288_decode_init(AVCodecContext *avctx) |
| 63 | { |
| 64 | RA288Context *ractx = avctx->priv_data; |
| 65 | |
| 66 | avctx->channels = 1; |
| 67 | avctx->channel_layout = AV_CH_LAYOUT_MONO; |
| 68 | avctx->sample_fmt = AV_SAMPLE_FMT_FLT; |
| 69 | |
| 70 | if (avctx->block_align <= 0) { |
| 71 | av_log(avctx, AV_LOG_ERROR, "unsupported block align\n"); |
| 72 | return AVERROR_PATCHWELCOME; |
| 73 | } |
| 74 | |
| 75 | avpriv_float_dsp_init(&ractx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT); |
| 76 | |
| 77 | return 0; |
| 78 | } |
| 79 | |
| 80 | static void convolve(float *tgt, const float *src, int len, int n) |
| 81 | { |
| 82 | for (; n >= 0; n--) |
| 83 | tgt[n] = avpriv_scalarproduct_float_c(src, src - n, len); |
| 84 | |
| 85 | } |
| 86 | |
| 87 | static void decode(RA288Context *ractx, float gain, int cb_coef) |
| 88 | { |
| 89 | int i; |
| 90 | double sumsum; |
| 91 | float sum, buffer[5]; |
| 92 | float *block = ractx->sp_hist + 70 + 36; // current block |
| 93 | float *gain_block = ractx->gain_hist + 28; |
| 94 | |
| 95 | memmove(ractx->sp_hist + 70, ractx->sp_hist + 75, 36*sizeof(*block)); |
| 96 | |
| 97 | /* block 46 of G.728 spec */ |
| 98 | sum = 32.0; |
| 99 | for (i=0; i < 10; i++) |
| 100 | sum -= gain_block[9-i] * ractx->gain_lpc[i]; |
| 101 | |
| 102 | /* block 47 of G.728 spec */ |
| 103 | sum = av_clipf(sum, 0, 60); |
| 104 | |
| 105 | /* block 48 of G.728 spec */ |
| 106 | /* exp(sum * 0.1151292546497) == pow(10.0,sum/20) */ |
| 107 | sumsum = exp(sum * 0.1151292546497) * gain * (1.0/(1<<23)); |
| 108 | |
| 109 | for (i=0; i < 5; i++) |
| 110 | buffer[i] = codetable[cb_coef][i] * sumsum; |
| 111 | |
| 112 | sum = avpriv_scalarproduct_float_c(buffer, buffer, 5); |
| 113 | |
| 114 | sum = FFMAX(sum, 5.0 / (1<<24)); |
| 115 | |
| 116 | /* shift and store */ |
| 117 | memmove(gain_block, gain_block + 1, 9 * sizeof(*gain_block)); |
| 118 | |
| 119 | gain_block[9] = 10 * log10(sum) + (10*log10(((1<<24)/5.)) - 32); |
| 120 | |
| 121 | ff_celp_lp_synthesis_filterf(block, ractx->sp_lpc, buffer, 5, 36); |
| 122 | } |
| 123 | |
| 124 | /** |
| 125 | * Hybrid window filtering, see blocks 36 and 49 of the G.728 specification. |
| 126 | * |
| 127 | * @param order filter order |
| 128 | * @param n input length |
| 129 | * @param non_rec number of non-recursive samples |
| 130 | * @param out filter output |
| 131 | * @param hist pointer to the input history of the filter |
| 132 | * @param out pointer to the non-recursive part of the output |
| 133 | * @param out2 pointer to the recursive part of the output |
| 134 | * @param window pointer to the windowing function table |
| 135 | */ |
| 136 | static void do_hybrid_window(RA288Context *ractx, |
| 137 | int order, int n, int non_rec, float *out, |
| 138 | float *hist, float *out2, const float *window) |
| 139 | { |
| 140 | int i; |
| 141 | float buffer1[MAX_BACKWARD_FILTER_ORDER + 1]; |
| 142 | float buffer2[MAX_BACKWARD_FILTER_ORDER + 1]; |
| 143 | LOCAL_ALIGNED(32, float, work, [FFALIGN(MAX_BACKWARD_FILTER_ORDER + |
| 144 | MAX_BACKWARD_FILTER_LEN + |
| 145 | MAX_BACKWARD_FILTER_NONREC, 16)]); |
| 146 | |
| 147 | av_assert2(order>=0); |
| 148 | |
| 149 | ractx->fdsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16)); |
| 150 | |
| 151 | convolve(buffer1, work + order , n , order); |
| 152 | convolve(buffer2, work + order + n, non_rec, order); |
| 153 | |
| 154 | for (i=0; i <= order; i++) { |
| 155 | out2[i] = out2[i] * 0.5625 + buffer1[i]; |
| 156 | out [i] = out2[i] + buffer2[i]; |
| 157 | } |
| 158 | |
| 159 | /* Multiply by the white noise correcting factor (WNCF). */ |
| 160 | *out *= 257.0 / 256.0; |
| 161 | } |
| 162 | |
| 163 | /** |
| 164 | * Backward synthesis filter, find the LPC coefficients from past speech data. |
| 165 | */ |
| 166 | static void backward_filter(RA288Context *ractx, |
| 167 | float *hist, float *rec, const float *window, |
| 168 | float *lpc, const float *tab, |
| 169 | int order, int n, int non_rec, int move_size) |
| 170 | { |
| 171 | float temp[MAX_BACKWARD_FILTER_ORDER+1]; |
| 172 | |
| 173 | do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window); |
| 174 | |
| 175 | if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1)) |
| 176 | ractx->fdsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 16)); |
| 177 | |
| 178 | memmove(hist, hist + n, move_size*sizeof(*hist)); |
| 179 | } |
| 180 | |
| 181 | static int ra288_decode_frame(AVCodecContext * avctx, void *data, |
| 182 | int *got_frame_ptr, AVPacket *avpkt) |
| 183 | { |
| 184 | AVFrame *frame = data; |
| 185 | const uint8_t *buf = avpkt->data; |
| 186 | int buf_size = avpkt->size; |
| 187 | float *out; |
| 188 | int i, ret; |
| 189 | RA288Context *ractx = avctx->priv_data; |
| 190 | GetBitContext gb; |
| 191 | |
| 192 | if (buf_size < avctx->block_align) { |
| 193 | av_log(avctx, AV_LOG_ERROR, |
| 194 | "Error! Input buffer is too small [%d<%d]\n", |
| 195 | buf_size, avctx->block_align); |
| 196 | return AVERROR_INVALIDDATA; |
| 197 | } |
| 198 | |
| 199 | /* get output buffer */ |
| 200 | frame->nb_samples = RA288_BLOCK_SIZE * RA288_BLOCKS_PER_FRAME; |
| 201 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
| 202 | return ret; |
| 203 | out = (float *)frame->data[0]; |
| 204 | |
| 205 | init_get_bits8(&gb, buf, avctx->block_align); |
| 206 | |
| 207 | for (i=0; i < RA288_BLOCKS_PER_FRAME; i++) { |
| 208 | float gain = amptable[get_bits(&gb, 3)]; |
| 209 | int cb_coef = get_bits(&gb, 6 + (i&1)); |
| 210 | |
| 211 | decode(ractx, gain, cb_coef); |
| 212 | |
| 213 | memcpy(out, &ractx->sp_hist[70 + 36], RA288_BLOCK_SIZE * sizeof(*out)); |
| 214 | out += RA288_BLOCK_SIZE; |
| 215 | |
| 216 | if ((i & 7) == 3) { |
| 217 | backward_filter(ractx, ractx->sp_hist, ractx->sp_rec, syn_window, |
| 218 | ractx->sp_lpc, syn_bw_tab, 36, 40, 35, 70); |
| 219 | |
| 220 | backward_filter(ractx, ractx->gain_hist, ractx->gain_rec, gain_window, |
| 221 | ractx->gain_lpc, gain_bw_tab, 10, 8, 20, 28); |
| 222 | } |
| 223 | } |
| 224 | |
| 225 | *got_frame_ptr = 1; |
| 226 | |
| 227 | return avctx->block_align; |
| 228 | } |
| 229 | |
| 230 | AVCodec ff_ra_288_decoder = { |
| 231 | .name = "real_288", |
| 232 | .long_name = NULL_IF_CONFIG_SMALL("RealAudio 2.0 (28.8K)"), |
| 233 | .type = AVMEDIA_TYPE_AUDIO, |
| 234 | .id = AV_CODEC_ID_RA_288, |
| 235 | .priv_data_size = sizeof(RA288Context), |
| 236 | .init = ra288_decode_init, |
| 237 | .decode = ra288_decode_frame, |
| 238 | .capabilities = CODEC_CAP_DR1, |
| 239 | }; |