summaryrefslogtreecommitdiff
path: root/libavcodec/truespeech.c
blob: 3f56973e0e57b66a9ba6b2b7bb6094b82e395c18 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
/*
 * DSP Group TrueSpeech compatible decoder
 * Copyright (c) 2005 Konstantin Shishkov
 *
 * This file is part of Libav.
 *
 * 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.
 *
 * 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 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 "dsputil.h"
#include "get_bits.h"
#include "internal.h"

#include "truespeech_data.h"
/**
 * @file
 * TrueSpeech decoder.
 */

/**
 * TrueSpeech decoder context
 */
typedef struct {
    DSPContext dsp;
    /* input data */
    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
    int16_t tmp1[8];    // coefficients for adding to out
    int16_t tmp2[8];    // coefficients for adding to out
    int16_t tmp3[8];    // coefficients for adding to out
    int16_t cvector[8]; // correlated input vector
    int filtval;        // gain value for one function
    int16_t newvec[60]; // tmp vector
    int16_t filters[32]; // filters for every subframe
} TSContext;

static av_cold int truespeech_decode_init(AVCodecContext * avctx)
{
    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_dsputil_init(&c->dsp, avctx);

    return 0;
}

static void truespeech_read_frame(TSContext *dec, const uint8_t *input)
{
    GetBitContext gb;

    dec->dsp.bswap_buf((uint32_t *)dec->buffer, (const uint32_t *)input, 8);
    init_get_bits(&gb, dec->buffer, 32 * 8);

    dec->vector[7] = ts_codebook[7][get_bits(&gb, 3)];
    dec->vector[6] = ts_codebook[6][get_bits(&gb, 3)];
    dec->vector[5] = ts_codebook[5][get_bits(&gb, 3)];
    dec->vector[4] = ts_codebook[4][get_bits(&gb, 4)];
    dec->vector[3] = ts_codebook[3][get_bits(&gb, 4)];
    dec->vector[2] = ts_codebook[2][get_bits(&gb, 4)];
    dec->vector[1] = ts_codebook[1][get_bits(&gb, 5)];
    dec->vector[0] = ts_codebook[0][get_bits(&gb, 5)];
    dec->flag      = get_bits1(&gb);

    dec->offset1[0] = get_bits(&gb, 4) << 4;
    dec->offset2[3] = get_bits(&gb, 7);
    dec->offset2[2] = get_bits(&gb, 7);
    dec->offset2[1] = get_bits(&gb, 7);
    dec->offset2[0] = get_bits(&gb, 7);

    dec->offset1[1]  = get_bits(&gb, 4);
    dec->pulseval[1] = get_bits(&gb, 14);
    dec->pulseval[0] = get_bits(&gb, 14);

    dec->offset1[1] |= get_bits(&gb, 4) << 4;
    dec->pulseval[3] = get_bits(&gb, 14);
    dec->pulseval[2] = get_bits(&gb, 14);

    dec->offset1[0] |= get_bits1(&gb);
    dec->pulsepos[0] = get_bits_long(&gb, 27);
    dec->pulseoff[0] = get_bits(&gb, 4);

    dec->offset1[0] |= get_bits1(&gb) << 1;
    dec->pulsepos[1] = get_bits_long(&gb, 27);
    dec->pulseoff[1] = get_bits(&gb, 4);

    dec->offset1[0] |= get_bits1(&gb) << 2;
    dec->pulsepos[2] = get_bits_long(&gb, 27);
    dec->pulseoff[2] = get_bits(&gb, 4);

    dec->offset1[0] |= get_bits1(&gb) << 3;
    dec->pulsepos[3] = get_bits_long(&gb, 27);
    dec->pulseoff[3] = get_bits(&gb, 4);
}

static void truespeech_correlate_filter(TSContext *dec)
{
    int16_t tmp[8];
    int i, j;

    for(i = 0; i < 8; i++){
        if(i > 0){
            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_decay_994_1000[i]) >> 15;

    dec->filtval = dec->vector[0];
}

static void truespeech_filters_merge(TSContext *dec)
{
    int i;

    if(!dec->flag){
        for(i = 0; i < 8; i++){
            dec->filters[i + 0] = dec->prevfilt[i];
            dec->filters[i + 8] = dec->prevfilt[i];
        }
    }else{
        for(i = 0; i < 8; i++){
            dec->filters[i + 0]=(dec->cvector[i] * 21846 + dec->prevfilt[i] * 10923 + 16384) >> 15;
            dec->filters[i + 8]=(dec->cvector[i] * 10923 + dec->prevfilt[i] * 21846 + 16384) >> 15;
        }
    }
    for(i = 0; i < 8; i++){
        dec->filters[i + 16] = dec->cvector[i];
        dec->filters[i + 24] = dec->cvector[i];
    }
}

static void truespeech_apply_twopoint_filter(TSContext *dec, int quart)
{
    int16_t tmp[146 + 60], *ptr0, *ptr1;
    const int16_t *filter;
    int i, t, off;

    t = dec->offset2[quart];
    if(t == 127){
        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 = ts_order2_coeffs + (t % 25) * 2;
    for(i = 0; i < 60; i++){
        t = (ptr0[0] * filter[0] + ptr0[1] * filter[1] + 0x2000) >> 14;
        ptr0++;
        dec->newvec[i] = t;
        ptr1[i] = t;
    }
}

static void truespeech_place_pulses(TSContext *dec, int16_t *out, int quart)
{
    int16_t tmp[7];
    int i, j, t;
    const int16_t *ptr1;
    int16_t *ptr2;
    int coef;

    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_pulse_scales[dec->pulseoff[quart] * 4 + t];
    }

    coef = dec->pulsepos[quart] >> 15;
    ptr1 = ts_pulse_values + 30;
    ptr2 = tmp;
    for(i = 0, j = 3; (i < 30) && (j > 0); i++){
        t = *ptr1++;
        if(coef >= t)
            coef -= t;
        else{
            out[i] = *ptr2++;
            ptr1 += 30;
            j--;
        }
    }
    coef = dec->pulsepos[quart] & 0x7FFF;
    ptr1 = ts_pulse_values;
    for(i = 30, j = 4; (i < 60) && (j > 0); i++){
        t = *ptr1++;
        if(coef >= t)
            coef -= t;
        else{
            out[i] = *ptr2++;
            ptr1 += 30;
            j--;
        }
    }

}

static void truespeech_update_filters(TSContext *dec, int16_t *out, int quart)
{
    int i;

    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];
    }
}

static void truespeech_synth(TSContext *dec, int16_t *out, int quart)
{
    int i,k;
    int t[8];
    int16_t *ptr0, *ptr1;

    ptr0 = dec->tmp1;
    ptr1 = dec->filters + quart * 8;
    for(i = 0; i < 60; i++){
        int sum = 0;
        for(k = 0; k < 8; k++)
            sum += ptr0[k] * ptr1[k];
        sum = (sum + (out[i] << 12) + 0x800) >> 12;
        out[i] = av_clip(sum, -0x7FFE, 0x7FFE);
        for(k = 7; k > 0; k--)
            ptr0[k] = ptr0[k - 1];
        ptr0[0] = out[i];
    }

    for(i = 0; i < 8; i++)
        t[i] = (ts_decay_35_64[i] * ptr1[i]) >> 15;

    ptr0 = dec->tmp2;
    for(i = 0; i < 60; i++){
        int sum = 0;
        for(k = 0; k < 8; k++)
            sum += ptr0[k] * t[k];
        for(k = 7; k > 0; k--)
            ptr0[k] = ptr0[k - 1];
        ptr0[0] = out[i];
        out[i] = ((out[i] << 12) - sum) >> 12;
    }

    for(i = 0; i < 8; i++)
        t[i] = (ts_decay_3_4[i] * ptr1[i]) >> 15;

    ptr0 = dec->tmp3;
    for(i = 0; i < 60; i++){
        int sum = out[i] << 12;
        for(k = 0; k < 8; k++)
            sum += ptr0[k] * t[k];
        for(k = 7; k > 0; k--)
            ptr0[k] = ptr0[k - 1];
        ptr0[0] = av_clip((sum + 0x800) >> 12, -0x7FFE, 0x7FFE);

        sum = ((ptr0[1] * (dec->filtval - (dec->filtval >> 2))) >> 4) + sum;
        sum = sum - (sum >> 3);
        out[i] = av_clip((sum + 0x800) >> 12, -0x7FFE, 0x7FFE);
    }
}

static void truespeech_save_prevvec(TSContext *c)
{
    int i;

    for(i = 0; i < 8; i++)
        c->prevfilt[i] = c->cvector[i];
}

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;
    int16_t *samples;
    int iterations, ret;

    iterations = buf_size / 32;

    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));

    for(j = 0; j < iterations; j++) {
        truespeech_read_frame(c, buf);
        buf += 32;

        truespeech_correlate_filter(c);
        truespeech_filters_merge(c);

        for(i = 0; i < 4; i++) {
            truespeech_apply_twopoint_filter(c, i);
            truespeech_place_pulses  (c, samples, i);
            truespeech_update_filters(c, samples, i);
            truespeech_synth         (c, samples, i);
            samples += 60;
        }

        truespeech_save_prevvec(c);
    }

    *got_frame_ptr = 1;

    return buf_size;
}

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   = CODEC_CAP_DR1,
};