FAAD是比较成熟高效的开源AAC解码库,这里用于解码AAC生成PCM数据,用于音频播放。这里因为faad库,会将单通道转化为双通道踩了些坑,所以记录一下。
我使用的是2.11.0版本,貌似往前的版本没有使用CMake,需要configure配置编译
1.源码编译
使用git拉取
git clone https://github.com/knik0/faad2.git
因为是交叉编译,所以创建了一个cfg_file_path,其中设置一些参数,cfg_file_path内容如下:
##################################
# 配置 交叉编译
#################################
set(CMAKE_SYSTEM_NAME Linux) #设置目标系统名字
set(CMAKE_SYSTEM_PROCESSOR aarch64) #设置目标处理器架构
set(CMAKE_C_COMPILER aarch64-ca53-linux-gnu-gcc-10.4.0)#设置交叉编译器
add_compile_definitions(DISABLE_SBR) #禁止SBR和DECODER
add_compile_definitions(LC_ONLY_DECODER)
指定cfg_file_path进行编译,编译产物输出到output目录
mkdir output
/usr/bin/cmake -DCMAKE_TOOLCHAIN_FILE=cfg_file_path -Boutput/
cd output
make
sudo make install
这样编译完成后在output下就会有libfaad.so,libfaad.so.2,libfaad.so.2.11.1,可以拷贝到开发板即可
2.测试代码
/**
* faaddec.c
* use faad library to decode AAC, only can decode frame with ADTS head
*/
#include <stdio.h>
#include <memory.h>
#include "faad.h"
#define FRAME_MAX_LEN 1024*5
#define BUFFER_MAX_LEN 1024*1024
void show_usage()
{
printf("usage\nfaaddec src_file dst_file");
}
/**
* fetch one ADTS frame
*/
int get_one_ADTS_frame(unsigned char* buffer, size_t buf_size, unsigned char* data ,size_t* data_size)
{
size_t size = 0;
if(!buffer || !data || !data_size )
{
return -1;
}
while(1)
{
if(buf_size < 7 )
{
return -1;
}
if((buffer[0] == 0xff) && ((buffer[1] & 0xf0) == 0xf0) )
{
size |= ((buffer[3] & 0x03) <<11); //high 2 bit
size |= buffer[4]<<3; //middle 8 bit
size |= ((buffer[5] & 0xe0)>>5); //low 3bit
break;
}
--buf_size;
++buffer;
}
if(buf_size < size)
{
return -1;
}
memcpy(data, buffer, size);
*data_size = size;
return 0;
}
int main(int argc, char* argv[])
{
static unsigned char frame[FRAME_MAX_LEN];
static unsigned char buffer[BUFFER_MAX_LEN] = {0};
char src_file[128] = {0};
char dst_file[128] = {0};
FILE* ifile = NULL;
FILE* ofile = NULL;
unsigned long samplerate;
unsigned char channels;
NeAACDecHandle decoder = 0;
size_t data_size = 0;
size_t size = 0;
NeAACDecFrameInfo frame_info;
unsigned char* input_data = buffer;
unsigned char* pcm_data = NULL;
//analyse parameter
if(argc < 3)
{
show_usage();
return -1;
}
sscanf(argv[1], "%s", src_file);
sscanf(argv[2], "%s", dst_file);
ifile = fopen(src_file, "rb");
ofile = fopen(dst_file, "wb");
if(!ifile || !ofile)
{
printf("source or destination file");
return -1;
}
data_size = fread(buffer, 1, BUFFER_MAX_LEN, ifile);
//open decoder
decoder = NeAACDecOpen();
if(get_one_ADTS_frame(buffer, data_size, frame, &size) < 0)
{
return -1;
}
NeAACDecConfigurationPtr config = NeAACDecGetCurrentConfiguration(decoder);
config->defObjectType = LC;
config->defSampleRate = 8000;
// config->defObjectType = HE_AAC;
config->outputFormat = FAAD_FMT_16BIT; //位深度设置成16bit
config->downMatrix = 0;
// 源码部分如下:
// if (*samplerate <= 24000 && (hDecoder->config.dontUpSampleImplicitSBR == 0))
// {
// *samplerate *= 2;
// hDecoder->forceUpSampling = 1;
// } else if (*samplerate > 24000 && (hDecoder->config.dontUpSampleImplicitSBR == 0)) {
// hDecoder->downSampledSBR = 1;
// }
// dontUpSampleImplicitSBR 设置成1,可以禁止NeAACDecInit时,采样率(小于等于24000时)翻倍的问题
config->dontUpSampleImplicitSBR = 1;
NeAACDecSetConfiguration(decoder, config);
//initialize decoder
NeAACDecInit(decoder, frame, size, &samplerate, &channels);
printf("samplerate %d, channels %d\n", samplerate, channels);
while(get_one_ADTS_frame(input_data, data_size, frame, &size) == 0)
{
// printf("frame size %d\n", size);
//decode ADTS frame
pcm_data = (unsigned char*)NeAACDecDecode(decoder, &frame_info, frame, size);
if(frame_info.error > 0)
{
printf("%s\n",NeAACDecGetErrorMessage(frame_info.error));
}
else if(pcm_data && frame_info.samples > 0)
{
printf("frame info: bytesconsumed %d, channels %d, header_type %d\
object_type %d, samples %d, samplerate %d\n",
frame_info.bytesconsumed,
frame_info.channels, frame_info.header_type,
frame_info.object_type, frame_info.samples,
frame_info.samplerate);
fwrite(pcm_data, 1, frame_info.samples * frame_info.channels * 2, ofile); //2个通道
fflush(ofile);
}
data_size -= size;
input_data += size;
}
NeAACDecClose(decoder);
fclose(ifile);
fclose(ofile);
return 0;
}
注意这里很多博客给的都是错的,frame_info.samples * frame_info.channels * 2,这里需要乘以2,因为位宽是16bit,而一个字节是8bit。在使用单通道的情况下,这里如果不乘以2,会导致音频加速等问题。
3.程序运行
使用下面的编译指令编译,拷贝可执行文件和动态库到板子里。
aarch64-ca53-linux-gnu-gcc-10.4.0 main.c -o main -lfaad -L/usr/local/lib/
./main_aarch high_temperature_shut_down.aac test.pcm
动态库拷贝到板子的/usr/lib下后,在板子上运行可执行文件。
4.关于单/双通道
faad这个在内部因为一些音频算法,开启一些编译选项之后,会将通道数乘以2,导致输出的音频通道数总是2,达不到预期效果,要解决这个问题,需要做以下两点
1)参考第二点测试代码,需要乘以2,否则音频加速
2)参考第一点中的编译配置选项,取消一些编译配置选项,防止通道乘以2