* DSP Group TrueSpeech compatible decoder
* Copyright (c) 2005 Konstantin Shishkov
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+
+#include "libavutil/channel_layout.h"
+#include "libavutil/intreadwrite.h"
+
#include "avcodec.h"
+#include "bitstream.h"
+#include "bswapdsp.h"
+#include "internal.h"
#include "truespeech_data.h"
+
/**
- * @file truespeech.c
+ * @file
* TrueSpeech decoder.
*/
/**
* TrueSpeech decoder context
*/
-typedef struct {
+typedef struct TSContext {
+ BswapDSPContext bdsp;
/* input data */
- int16_t vector[8]; //< input vector: 5/5/4/4/4/3/3/3
- int offset1[2]; //< 8-bit value, used in one copying offset
- int offset2[4]; //< 7-bit value, encodes offsets for copying and for two-point filter
- int pulseoff[4]; //< 4-bit offset of pulse values block
- int pulsepos[4]; //< 27-bit variable, encodes 7 pulse positions
- int pulseval[4]; //< 7x2-bit pulse values
- int flag; //< 1-bit flag, shows how to choose filters
+ DECLARE_ALIGNED(16, uint8_t, buffer)[32];
+ int16_t vector[8]; ///< input vector: 5/5/4/4/4/3/3/3
+ int offset1[2]; ///< 8-bit value, used in one copying offset
+ int offset2[4]; ///< 7-bit value, encodes offsets for copying and for two-point filter
+ int pulseoff[4]; ///< 4-bit offset of pulse values block
+ int pulsepos[4]; ///< 27-bit variable, encodes 7 pulse positions
+ int pulseval[4]; ///< 7x2-bit pulse values
+ int flag; ///< 1-bit flag, shows how to choose filters
/* temporary data */
int filtbuf[146]; // some big vector used for storing filters
int prevfilt[8]; // filter from previous frame
static av_cold int truespeech_decode_init(AVCodecContext * avctx)
{
-// TSContext *c = avctx->priv_data;
+ TSContext *c = avctx->priv_data;
+
+ if (avctx->channels != 1) {
+ avpriv_request_sample(avctx, "Channel count %d", avctx->channels);
+ return AVERROR_PATCHWELCOME;
+ }
+
+ avctx->channel_layout = AV_CH_LAYOUT_MONO;
+ avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+
+ ff_bswapdsp_init(&c->bdsp);
return 0;
}
static void truespeech_read_frame(TSContext *dec, const uint8_t *input)
{
- uint32_t t;
-
- /* first dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->flag = t & 1;
-
- dec->vector[0] = ts_codebook[0][(t >> 1) & 0x1F];
- dec->vector[1] = ts_codebook[1][(t >> 6) & 0x1F];
- dec->vector[2] = ts_codebook[2][(t >> 11) & 0xF];
- dec->vector[3] = ts_codebook[3][(t >> 15) & 0xF];
- dec->vector[4] = ts_codebook[4][(t >> 19) & 0xF];
- dec->vector[5] = ts_codebook[5][(t >> 23) & 0x7];
- dec->vector[6] = ts_codebook[6][(t >> 26) & 0x7];
- dec->vector[7] = ts_codebook[7][(t >> 29) & 0x7];
-
- /* second dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->offset2[0] = (t >> 0) & 0x7F;
- dec->offset2[1] = (t >> 7) & 0x7F;
- dec->offset2[2] = (t >> 14) & 0x7F;
- dec->offset2[3] = (t >> 21) & 0x7F;
-
- dec->offset1[0] = ((t >> 28) & 0xF) << 4;
-
- /* third dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulseval[0] = (t >> 0) & 0x3FFF;
- dec->pulseval[1] = (t >> 14) & 0x3FFF;
-
- dec->offset1[1] = (t >> 28) & 0x0F;
-
- /* fourth dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulseval[2] = (t >> 0) & 0x3FFF;
- dec->pulseval[3] = (t >> 14) & 0x3FFF;
-
- dec->offset1[1] |= ((t >> 28) & 0x0F) << 4;
-
- /* fifth dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulsepos[0] = (t >> 4) & 0x7FFFFFF;
-
- dec->pulseoff[0] = (t >> 0) & 0xF;
-
- dec->offset1[0] |= (t >> 31) & 1;
-
- /* sixth dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulsepos[1] = (t >> 4) & 0x7FFFFFF;
-
- dec->pulseoff[1] = (t >> 0) & 0xF;
-
- dec->offset1[0] |= ((t >> 31) & 1) << 1;
-
- /* seventh dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulsepos[2] = (t >> 4) & 0x7FFFFFF;
-
- dec->pulseoff[2] = (t >> 0) & 0xF;
-
- dec->offset1[0] |= ((t >> 31) & 1) << 2;
-
- /* eighth dword */
- t = AV_RL32(input);
- input += 4;
-
- dec->pulsepos[3] = (t >> 4) & 0x7FFFFFF;
-
- dec->pulseoff[3] = (t >> 0) & 0xF;
-
- dec->offset1[0] |= ((t >> 31) & 1) << 3;
-
+ BitstreamContext bc;
+
+ dec->bdsp.bswap_buf((uint32_t *) dec->buffer, (const uint32_t *) input, 8);
+ bitstream_init8(&bc, dec->buffer, 32);
+
+ dec->vector[7] = ts_codebook[7][bitstream_read(&bc, 3)];
+ dec->vector[6] = ts_codebook[6][bitstream_read(&bc, 3)];
+ dec->vector[5] = ts_codebook[5][bitstream_read(&bc, 3)];
+ dec->vector[4] = ts_codebook[4][bitstream_read(&bc, 4)];
+ dec->vector[3] = ts_codebook[3][bitstream_read(&bc, 4)];
+ dec->vector[2] = ts_codebook[2][bitstream_read(&bc, 4)];
+ dec->vector[1] = ts_codebook[1][bitstream_read(&bc, 5)];
+ dec->vector[0] = ts_codebook[0][bitstream_read(&bc, 5)];
+ dec->flag = bitstream_read_bit(&bc);
+
+ dec->offset1[0] = bitstream_read(&bc, 4) << 4;
+ dec->offset2[3] = bitstream_read(&bc, 7);
+ dec->offset2[2] = bitstream_read(&bc, 7);
+ dec->offset2[1] = bitstream_read(&bc, 7);
+ dec->offset2[0] = bitstream_read(&bc, 7);
+
+ dec->offset1[1] = bitstream_read(&bc, 4);
+ dec->pulseval[1] = bitstream_read(&bc, 14);
+ dec->pulseval[0] = bitstream_read(&bc, 14);
+
+ dec->offset1[1] |= bitstream_read(&bc, 4) << 4;
+ dec->pulseval[3] = bitstream_read(&bc, 14);
+ dec->pulseval[2] = bitstream_read(&bc, 14);
+
+ dec->offset1[0] |= bitstream_read_bit(&bc);
+ dec->pulsepos[0] = bitstream_read(&bc, 27);
+ dec->pulseoff[0] = bitstream_read(&bc, 4);
+
+ dec->offset1[0] |= bitstream_read_bit(&bc) << 1;
+ dec->pulsepos[1] = bitstream_read(&bc, 27);
+ dec->pulseoff[1] = bitstream_read(&bc, 4);
+
+ dec->offset1[0] |= bitstream_read_bit(&bc) << 2;
+ dec->pulsepos[2] = bitstream_read(&bc, 27);
+ dec->pulseoff[2] = bitstream_read(&bc, 4);
+
+ dec->offset1[0] |= bitstream_read_bit(&bc) << 3;
+ dec->pulsepos[3] = bitstream_read(&bc, 27);
+ dec->pulseoff[3] = bitstream_read(&bc, 4);
}
static void truespeech_correlate_filter(TSContext *dec)
for(i = 0; i < 8; i++){
if(i > 0){
- memcpy(tmp, dec->cvector, i * 2);
+ memcpy(tmp, dec->cvector, i * sizeof(*tmp));
for(j = 0; j < i; j++)
dec->cvector[j] = ((tmp[i - j - 1] * dec->vector[i]) +
(dec->cvector[j] << 15) + 0x4000) >> 15;
dec->cvector[i] = (8 - dec->vector[i]) >> 3;
}
for(i = 0; i < 8; i++)
- dec->cvector[i] = (dec->cvector[i] * ts_230[i]) >> 15;
+ dec->cvector[i] = (dec->cvector[i] * ts_decay_994_1000[i]) >> 15;
dec->filtval = dec->vector[0];
}
t = dec->offset2[quart];
if(t == 127){
- memset(dec->newvec, 0, 60 * 2);
+ memset(dec->newvec, 0, 60 * sizeof(*dec->newvec));
return;
}
for(i = 0; i < 146; i++)
tmp[i] = dec->filtbuf[i];
off = (t / 25) + dec->offset1[quart >> 1] + 18;
+ off = av_clip(off, 0, 145);
ptr0 = tmp + 145 - off;
ptr1 = tmp + 146;
- filter = (const int16_t*)ts_240 + (t % 25) * 2;
+ filter = ts_order2_coeffs + (t % 25) * 2;
for(i = 0; i < 60; i++){
t = (ptr0[0] * filter[0] + ptr0[1] * filter[1] + 0x2000) >> 14;
ptr0++;
int16_t *ptr2;
int coef;
- memset(out, 0, 60 * 2);
+ memset(out, 0, 60 * sizeof(*out));
for(i = 0; i < 7; i++) {
t = dec->pulseval[quart] & 3;
dec->pulseval[quart] >>= 2;
- tmp[6 - i] = ts_562[dec->pulseoff[quart] * 4 + t];
+ tmp[6 - i] = ts_pulse_scales[dec->pulseoff[quart] * 4 + t];
}
coef = dec->pulsepos[quart] >> 15;
- ptr1 = (const int16_t*)ts_140 + 30;
+ ptr1 = ts_pulse_values + 30;
ptr2 = tmp;
for(i = 0, j = 3; (i < 30) && (j > 0); i++){
t = *ptr1++;
}
}
coef = dec->pulsepos[quart] & 0x7FFF;
- ptr1 = (const int16_t*)ts_140;
+ ptr1 = ts_pulse_values;
for(i = 30, j = 4; (i < 60) && (j > 0); i++){
t = *ptr1++;
if(coef >= t)
{
int i;
- for(i = 0; i < 86; i++)
- dec->filtbuf[i] = dec->filtbuf[i + 60];
+ memmove(dec->filtbuf, &dec->filtbuf[60], 86 * sizeof(*dec->filtbuf));
for(i = 0; i < 60; i++){
dec->filtbuf[i + 86] = out[i] + dec->newvec[i] - (dec->newvec[i] >> 3);
out[i] += dec->newvec[i];
}
for(i = 0; i < 8; i++)
- t[i] = (ts_5E2[i] * ptr1[i]) >> 15;
+ t[i] = (ts_decay_35_64[i] * ptr1[i]) >> 15;
ptr0 = dec->tmp2;
for(i = 0; i < 60; i++){
}
for(i = 0; i < 8; i++)
- t[i] = (ts_5F2[i] * ptr1[i]) >> 15;
+ t[i] = (ts_decay_3_4[i] * ptr1[i]) >> 15;
ptr0 = dec->tmp3;
for(i = 0; i < 60; i++){
c->prevfilt[i] = c->cvector[i];
}
-static int truespeech_decode_frame(AVCodecContext *avctx,
- void *data, int *data_size,
- const uint8_t *buf, int buf_size)
+static int truespeech_decode_frame(AVCodecContext *avctx, void *data,
+ int *got_frame_ptr, AVPacket *avpkt)
{
+ AVFrame *frame = data;
+ const uint8_t *buf = avpkt->data;
+ int buf_size = avpkt->size;
TSContext *c = avctx->priv_data;
int i, j;
- short *samples = data;
- int consumed = 0;
- int16_t out_buf[240];
- int iterations;
+ int16_t *samples;
+ int iterations, ret;
+
+ iterations = buf_size / 32;
- if (!buf_size)
- return 0;
+ if (!iterations) {
+ av_log(avctx, AV_LOG_ERROR,
+ "Too small input buffer (%d bytes), need at least 32 bytes\n", buf_size);
+ return -1;
+ }
+
+ /* get output buffer */
+ frame->nb_samples = iterations * 240;
+ if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+ return ret;
+ }
+ samples = (int16_t *)frame->data[0];
+
+ memset(samples, 0, iterations * 240 * sizeof(*samples));
- iterations = FFMIN(buf_size / 32, *data_size / 480);
for(j = 0; j < iterations; j++) {
- truespeech_read_frame(c, buf + consumed);
- consumed += 32;
+ truespeech_read_frame(c, buf);
+ buf += 32;
truespeech_correlate_filter(c);
truespeech_filters_merge(c);
- memset(out_buf, 0, 240 * 2);
for(i = 0; i < 4; i++) {
truespeech_apply_twopoint_filter(c, i);
- truespeech_place_pulses(c, out_buf + i * 60, i);
- truespeech_update_filters(c, out_buf + i * 60, i);
- truespeech_synth(c, out_buf + i * 60, i);
+ truespeech_place_pulses (c, samples, i);
+ truespeech_update_filters(c, samples, i);
+ truespeech_synth (c, samples, i);
+ samples += 60;
}
truespeech_save_prevvec(c);
-
- /* finally output decoded frame */
- for(i = 0; i < 240; i++)
- *samples++ = out_buf[i];
-
}
- *data_size = consumed * 15;
+ *got_frame_ptr = 1;
- return consumed;
+ return buf_size;
}
-AVCodec truespeech_decoder = {
- "truespeech",
- CODEC_TYPE_AUDIO,
- CODEC_ID_TRUESPEECH,
- sizeof(TSContext),
- truespeech_decode_init,
- NULL,
- NULL,
- truespeech_decode_frame,
- .long_name = NULL_IF_CONFIG_SMALL("DSP Group TrueSpeech"),
+AVCodec ff_truespeech_decoder = {
+ .name = "truespeech",
+ .long_name = NULL_IF_CONFIG_SMALL("DSP Group TrueSpeech"),
+ .type = AVMEDIA_TYPE_AUDIO,
+ .id = AV_CODEC_ID_TRUESPEECH,
+ .priv_data_size = sizeof(TSContext),
+ .init = truespeech_decode_init,
+ .decode = truespeech_decode_frame,
+ .capabilities = AV_CODEC_CAP_DR1,
};