MPEG-2音频解码

MPEG-2音频的解码过程：

1、检查4个字节的音频头，过程为：

static inline int ff_mpa_check_header(uint32_t header){
/* header */
if ((header & 0xffe00000) != 0xffe00000) //即前11位为固定头信息，全为1
return -1;
/* layer check */
if ((header & (3<<17)) == 0)//即第14、15两位表示layer
return -1;
/* bit rate */
if ((header & (0xf<<12)) == 0xf<<12)//即第17、18、19、20四位不能全为1
return -1;
/* frequency */
if ((header & (3<<10)) == 3<<10)//即第21、22位不能全为1
return -1;
return 0;
}

2、根据header信息得到相应参数信息：

s->layer = 4 - ((header >> 17) & 3);)//即第14、15两位与0x0011相与，再被4减

sample_rate_index = (header >> 10) & 3;//获取的是采样率的索引，ff_mpa_freq_tab[3] = { 44100, 48000, 32000 };

sample_rate = ff_mpa_freq_tab[sample_rate_index] >> (s->lsf + mpeg25);//其中，if (header & (1<<20))

//{s->lsf =(header & (1<<19)) ? 0 : 1;
// mpeg25 = 0;
// } else {
//s->lsf = 1;
//mpeg25 = 1;
// }

sample_rate_index += 3 * (s->lsf + mpeg25);
s->sample_rate_index = sample_rate_index;

s->sample_rate = sample_rate;

bitrate_index = (header >> 12) & 0xf;

s->mode = (header >> 6) & 3;

padding = (header >> 9) & 1;

if (s->mode == MPA_MONO)
s->nb_channels = 1;
else
s->nb_channels = 2;

if (bitrate_index != 0) {
frame_size = ff_mpa_bitrate_tab[s->lsf][s->layer - 1][bitrate_index];
s->bit_rate = frame_size * 1000;
switch(s->layer) {
case 1:
frame_size = (frame_size * 12000) / sample_rate;
frame_size = (frame_size + padding) * 4;
break;
case 2:
frame_size = (frame_size * 144000) / sample_rate;
frame_size += padding;
break;
default:
case 3:
frame_size = (frame_size * 144000) / (sample
fa50
_rate << s->lsf);
frame_size += padding;
break;
}
s->frame_size = frame_size;
} else {
/* if no frame size computed, signal it */
return 1;
}

其中，

const uint16_t ff_mpa_bitrate_tab[2][3][15] = {
{ {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448 },
{0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384 },
{0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320 } },
{ {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256},
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160},
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160}
}
};

根据layer数，即可知道该音频的编码类型：

layer = 1, 对应MPEG1

layer = 2, 对应MPEG2，framesize = 1152 //1152与外面开辟的保存输出数据的缓冲大小有关，如果开辟空间大小<1152*avctx- //>channels*2时，则空间太小；
layer = 3, 对应MPEG3（MP3）

通过比较送入的数据buf_size和解析header获取的数据大小 s->frame_size相比，即可得知当前送入的数据是否有问题：

如果s->frame_size > buf_size，则送入的帧不够；

如果s->frame_size < buf_size，则该帧可能有问题；

3、解码

3.1 /*先根据参数（bitrate、channel、samplerate）等信息，选取码表*/

table = ff_mpa_l2_select_table(s->bit_rate / 1000, s->nb_channels, s->sample_rate, s->lsf);

3.2 /* parse bit allocation */

j = 0;
for(i=0;i<bound;i++) {
bit_alloc_bits = alloc_table[j];
for(ch=0;ch<s->nb_channels;ch++) {
bit_alloc[ch][i] = get_bits(&s->gb, bit_alloc_bits);
}
j += 1 << bit_alloc_bits;
}
for(i=bound;i<sblimit;i++) {
bit_alloc_bits = alloc_table[j];
v = get_bits(&s->gb, bit_alloc_bits);
bit_alloc[0][i] = v;
bit_alloc[1][i] = v;
j += 1 << bit_alloc_bits;
}

3.3 /* scale codes */

for(i=0;i<sblimit;i++) {
for(ch=0;ch<s->nb_channels;ch++) {
if (bit_alloc[ch][i])
scale_code[ch][i] = get_bits(&s->gb, 2);
}
}

3.4 /* scale factors */

for(i=0;i<sblimit;i++) {
for(ch=0;ch<s->nb_channels;ch++) {
if (bit_alloc[ch][i]) {
sf = scale_factors[ch][i];
switch(scale_code[ch][i]) {
default:
case 0:
sf[0] = get_bits(&s->gb, 6);
sf[1] = get_bits(&s->gb, 6);
sf[2] = get_bits(&s->gb, 6);
break;
case 2:
sf[0] = get_bits(&s->gb, 6);
sf[1] = sf[0];
sf[2] = sf[0];
break;
case 1:
sf[0] = get_bits(&s->gb, 6);
sf[2] = get_bits(&s->gb, 6);
sf[1] = sf[0];
break;
case 3:
sf[0] = get_bits(&s->gb, 6);
sf[2] = get_bits(&s->gb, 6);
sf[1] = sf[2];
break;
}
}
}
}

3.5 /* samples */

for(k=0;k<3;k++) {
for(l=0;l<12;l+=3) {
j = 0;
for(i=0;i<bound;i++) {
bit_alloc_bits = alloc_table[j];
for(ch=0;ch<s->nb_channels;ch++) {
b = bit_alloc[ch][i];
if (b) {
scale = scale_factors[ch][i][k];
qindex = alloc_table[j+b];
bits = ff_mpa_quant_bits[qindex];
if (bits < 0) {
int v2;
/* 3 values at the same time */
v = get_bits(&s->gb, -bits);
v2 = division_tabs[qindex][v];
steps = ff_mpa_quant_steps[qindex];

s->sb_samples[ch][k * 12 + l + 0][i] =
l2_unscale_group(steps, v2 & 15, scale);
s->sb_samples[ch][k * 12 + l + 1][i] =
l2_unscale_group(steps, (v2 >> 4) & 15, scale);
s->sb_samples[ch][k * 12 + l + 2][i] =
l2_unscale_group(steps, v2 >> 8 , scale);
} else {
for(m=0;m<3;m++) {
v = get_bits(&s->gb, bits);
v = l1_unscale(bits - 1, v, scale);
s->sb_samples[ch][k * 12 + l + m][i] = v;
}
}
} else {
s->sb_samples[ch][k * 12 + l + 0][i] = 0;
s->sb_samples[ch][k * 12 + l + 1][i] = 0;
s->sb_samples[ch][k * 12 + l + 2][i] = 0;
}
}
/* next subband in alloc table */
j += 1 << bit_alloc_bits;
}
/* XXX: find a way to avoid this duplication of code */
for(i=bound;i<sblimit;i++) {
bit_alloc_bits = alloc_table[j];
b = bit_alloc[0][i];
if (b) {
int mant, scale0, scale1;
scale0 = scale_factors[0][i][k];
scale1 = scale_factors[1][i][k];
qindex = alloc_table[j+b];
bits = ff_mpa_quant_bits[qindex];
if (bits < 0) {
/* 3 values at the same time */
v = get_bits(&s->gb, -bits);
steps = ff_mpa_quant_steps[qindex];
mant = v % steps;
v = v / steps;
s->sb_samples[0][k * 12 + l + 0][i] =
l2_unscale_group(steps, mant, scale0);
s->sb_samples[1][k * 12 + l + 0][i] =
l2_unscale_group(steps, mant, scale1);
mant = v % steps;
v = v / steps;
s->sb_samples[0][k * 12 + l + 1][i] =
l2_unscale_group(steps, mant, scale0);
s->sb_samples[1][k * 12 + l + 1][i] =
l2_unscale_group(steps, mant, scale1);
s->sb_samples[0][k * 12 + l + 2][i] =
l2_unscale_group(steps, v, scale0);
s->sb_samples[1][k * 12 + l + 2][i] =
l2_unscale_group(steps, v, scale1);
} else {
for(m=0;m<3;m++) {
mant = get_bits(&s->gb, bits);
s->sb_samples[0][k * 12 + l + m][i] =
l1_unscale(bits - 1, mant, scale0);
s->sb_samples[1][k * 12 + l + m][i] =
l1_unscale(bits - 1, mant, scale1);
}
}
} else {
s->sb_samples[0][k * 12 + l + 0][i] = 0;
s->sb_samples[0][k * 12 + l + 1][i] = 0;
s->sb_samples[0][k * 12 + l + 2][i] = 0;
s->sb_samples[1][k * 12 + l + 0][i] = 0;
s->sb_samples[1][k * 12 + l + 1][i] = 0;
s->sb_samples[1][k * 12 + l + 2][i] = 0;
}
/* next subband in alloc table */
j += 1 << bit_alloc_bits;
}
/* fill remaining samples to zero */
for(i=sblimit;i<SBLIMIT;i++) {
for(ch=0;ch<s->nb_channels;ch++) {
s->sb_samples[ch][k * 12 + l + 0][i] = 0;
s->sb_samples[ch][k * 12 + l + 1][i] = 0;
s->sb_samples[ch][k * 12 + l + 2][i] = 0;
}
}
}
}

nb->frames = 36;

3.6 ff_mpa_synth_filter过程

先进行逆dct变换，dct32(synth_buf, sb_samples);

之后进行apply_window_mp3_c变换；

4.0 最后输出的PCM数据的实际大小为：

nb_frames * 32 * 2* s->nb_channels;