音频数据编解码——在.NET中使用Speex（附下载）
Speex是一套开源的音频编解码库，最新版本还包含了回音消除和防抖动等功能，如果我们想开发语音聊天或视频会议这样的系统，Speex将是一个不错的选择。到 http://www.speex.org可以下载Speex的源码（编译后的dll为libspeex.dll），最新版本为1.2。不过源码是用C++开发的，直接在.NET中使用会有诸多不便，为此，我用C#将其封装，使得编解码的调用相当简单。
由于Speex原始导出的API不是很方便C#调用，所以，在用C#封装之前，先要用C++对Speex的原始API进行简化，新建一个名为Speex的VC项目，然后引用libspeex.dll的相关库文件，添加cpp文件后，复制下列源码到文件中：
#include "speex\speex.h"
#include <windows.h>
#include <stdio.h>
#include <stdlib.h>
#include "speex/speex_echo.h"
#include "speex/speex_preprocess.h"
#include "Speex.h"
#define FRAME_SIZE 160
float encoder_input[FRAME_SIZE];
void *encoder_state;
SpeexBits encoder_bits;
BOOL APIENTRY DllMain( HANDLE hModule,
DWORD ul_reason_for_call,
LPVOID lpReserved
)
{
return TRUE;
}
#include <windows.h>
#include <stdio.h>
#include <stdlib.h>
#include "speex/speex_echo.h"
#include "speex/speex_preprocess.h"
#include "Speex.h"
#define FRAME_SIZE 160
float encoder_input[FRAME_SIZE];
void *encoder_state;
SpeexBits encoder_bits;
BOOL APIENTRY DllMain( HANDLE hModule,
DWORD ul_reason_for_call,
LPVOID lpReserved
)
{
return TRUE;
}
{
encoder_state = speex_encoder_init(&speex_nb_mode);
speex_encoder_ctl(encoder_state, SPEEX_SET_QUALITY, &quality);
speex_bits_init(&encoder_bits);
}
extern "C" __declspec(dllexport) void encoder_dispose()
{
speex_encoder_destroy(encoder_state);
speex_bits_destroy(&encoder_bits);
}
extern "C" __declspec(dllexport) int encoder_encode(const short *data, char *output)
{
for (int i = 0; i < FRAME_SIZE; i++)
encoder_input[i] = data[i];
speex_bits_reset(&encoder_bits);
speex_encode(encoder_state, encoder_input, &encoder_bits);
return speex_bits_write(&encoder_bits, output, 200);
}
float decoder_output[FRAME_SIZE];
void *decoder_state;
SpeexBits decoder_bits;
extern "C" __declspec(dllexport) void decoder_init()
{
decoder_state = speex_decoder_init(&speex_nb_mode);
int tmp = 1;
speex_decoder_ctl(decoder_state, SPEEX_SET_ENH, &tmp);
speex_bits_init(&decoder_bits);
}
extern "C" __declspec(dllexport) void decoder_dispose()
{
speex_decoder_destroy(decoder_state);
speex_bits_destroy(&decoder_bits);
}
extern "C" __declspec(dllexport) void decoder_decode(int nbBytes, char *data, short *output)
{
speex_bits_read_from(&decoder_bits, data, nbBytes);
speex_decode(decoder_state, &decoder_bits, decoder_output);
for (int i = 0; i < FRAME_SIZE; i++)
{
output[i] = decoder_output[i];
}
}
/*************************************************** 回音消除 **************************************/
bool m_bSpeexEchoHasInit;
SpeexEchoState* m_SpeexEchoState;
SpeexPreprocessState* m_pPreprocessorState;
int m_nFilterLen;
int m_nSampleRate;
float* m_pfNoise;
extern "C" __declspec(dllexport) void SpeexEchoCapture(short* input_frame, short* output_frame)
{
speex_echo_capture(m_SpeexEchoState, input_frame, output_frame);
}
extern "C" __declspec(dllexport) void SpeexEchoPlayback(short* echo_frame)
{
speex_echo_playback(m_SpeexEchoState, echo_frame);
}
extern "C" __declspec(dllexport) void SpeexEchoReset()
{
if (m_SpeexEchoState != NULL)
{
speex_echo_state_destroy(m_SpeexEchoState);
m_SpeexEchoState = NULL;
}
if (m_pPreprocessorState != NULL)
{
speex_preprocess_state_destroy(m_pPreprocessorState);
m_pPreprocessorState = NULL;
}
if (m_pfNoise != NULL)
{
delete []m_pfNoise;
m_pfNoise = NULL;
}
m_bSpeexEchoHasInit = false;
}
extern "C" __declspec(dllexport) void SpeexEchoInit(int filter_length, int sampling_rate ,bool associatePreprocesser)
{
SpeexEchoReset();
if (filter_length<=0 || sampling_rate<=0)
{
m_nFilterLen = 160*8;
m_nSampleRate = 8000;
}
else
{
m_nFilterLen = filter_length;
m_nSampleRate = sampling_rate;
}
m_SpeexEchoState = speex_echo_state_init(FRAME_SIZE, m_nFilterLen);
m_pPreprocessorState = speex_preprocess_state_init(FRAME_SIZE, m_nSampleRate);
if(associatePreprocesser)
{
speex_preprocess_ctl(m_pPreprocessorState, SPEEX_PREPROCESS_SET_ECHO_STATE,m_SpeexEchoState);
}
m_pfNoise = new float[FRAME_SIZE+1];
m_bSpeexEchoHasInit = true;
}
extern "C" __declspec(dllexport) void SpeexEchoDoAEC(short* mic, short* ref, short* out)
{
if (!m_bSpeexEchoHasInit)
{
return;
}
speex_echo_cancellation(m_SpeexEchoState,(const __int16 *) mic,(const __int16 *) ref,(__int16 *) out);
}
编译便生成Speex.dll。
如果对VC不熟悉也没关系，文末会直接给出libspeex.dll和Speex.dll的下载，直接使用就OK了。
现在，C#可以调用Speex.dll导出的简单函数了，最终封装的源码如下：
/// <summary>
/// 对Speex的C#封装。
/// zhuweisky 2010.05.13
/// </summary>
public class Speex :IAudioCodec
{
private const int FrameSize = 160;
#region IsDisposed
private volatile bool isDisposed = false;
public bool IsDisposed
{
get { return isDisposed; }
}
#endregion
#region Ctor
/// <summary>
/// 初始化。
/// </summary>
/// <param name="quality">编码质量，取值0~10</param>
public Speex(int quality)
{
if (quality < 0 || quality > 10)
{
throw new Exception("quality value must be between 0 and 10.");
}
Speex.encoder_init(quality);
Speex.decoder_init();
}
#endregion
#region Dispose
public void Dispose()
{
this.isDisposed = true;
System.Threading.Thread.Sleep(100);
Speex.decoder_dispose();
Speex.encoder_dispose();
}
#endregion
#region Encode
/// <summary>
/// 将采集到的音频数据进行编码。
/// </summary>
public byte[] Encode(byte[] data)
{
if (this.isDisposed)
{
return null;
}
if (data.Length % (FrameSize * 2) != 0)
{
throw new ArgumentException("Invalid Data Length.");
}
int nbBytes;
short[] input = new short[FrameSize];
byte[] buffer = new byte[200];
byte[] output = new byte[0];
for (int i = 0; i < data.Length / (FrameSize * 2); i++)
{
for (int j = 0; j < input.Length; j++)
{
input[j] = (short)(data[i * FrameSize * 2 + j * 2] + data[i * FrameSize * 2 + j * 2 + 1] * 0x100);
}
nbBytes = Speex.encoder_encode(input, buffer);
Array.Resize<byte>(ref output, output.Length + nbBytes + sizeof(int));
Array.Copy(buffer, 0, output, output.Length - nbBytes, nbBytes);
for (int j = 0; j < sizeof(int); j++)
{
output[output.Length - nbBytes - sizeof(int) + j] = (byte)(nbBytes % 0x100);
nbBytes /= 0x100;
}
}
return output;
}
#endregion
#region Decode
/// <summary>
/// 将编码后的数据进行解码得到原始的音频数据。
/// </summary>
public byte[] Decode(byte[] data)
{
if (this.isDisposed)
{
return null;
}
int nbBytes, index = 0;
byte[] input;
short[] buffer = new short[FrameSize];
byte[] output = new byte[0];
while (index < data.Length)
{
nbBytes = 0;
index += sizeof(int);
for (int i = 1; i <= sizeof(int); i++)
nbBytes = nbBytes * 0x100 + data[index - i];
input = new byte[nbBytes];
Array.Copy(data, index, input, 0, input.Length);
index += input.Length;
Speex.decoder_decode(nbBytes, input, buffer);
Array.Resize<byte>(ref output, output.Length + FrameSize * 2);
for (int i = 0; i < FrameSize; i++)
{
output[output.Length - FrameSize * 2 + i * 2] = (byte)(buffer[i] % 0x100);
output[output.Length - FrameSize * 2 + i * 2 + 1] = (byte)(buffer[i] / 0x100);
}
}
return output;
}
#endregion
#region Pinvoke
[DllImport("Speex.dll", EntryPoint = "encoder_init")]
internal extern static void encoder_init(int quality);
[DllImport("Speex.dll", EntryPoint = "encoder_dispose")]
internal extern static void encoder_dispose();
[DllImport("Speex.dll", EntryPoint = "encoder_encode")]
internal extern static int encoder_encode(short[] data, byte[] output);
[DllImport("Speex.dll", EntryPoint = "decoder_init")]
internal extern static void decoder_init();
[DllImport("Speex.dll", EntryPoint = "decoder_dispose")]
internal extern static void decoder_dispose();
[DllImport("Speex.dll", EntryPoint = "decoder_decode")]
internal extern static void decoder_decode(int nbBytes, byte[] data, short[] output);
#endregion
}
/// 对Speex的C#封装。
/// zhuweisky 2010.05.13
/// </summary>
public class Speex :IAudioCodec
{
private const int FrameSize = 160;
#region IsDisposed
private volatile bool isDisposed = false;
public bool IsDisposed
{
get { return isDisposed; }
}
#endregion
#region Ctor
/// <summary>
/// 初始化。
/// </summary>
/// <param name="quality">编码质量，取值0~10</param>
public Speex(int quality)
{
if (quality < 0 || quality > 10)
{
throw new Exception("quality value must be between 0 and 10.");
}
Speex.encoder_init(quality);
Speex.decoder_init();
}
#endregion
#region Dispose
public void Dispose()
{
this.isDisposed = true;
System.Threading.Thread.Sleep(100);
Speex.decoder_dispose();
Speex.encoder_dispose();
}
#endregion
#region Encode
/// <summary>
/// 将采集到的音频数据进行编码。
/// </summary>
public byte[] Encode(byte[] data)
{
if (this.isDisposed)
{
return null;
}
if (data.Length % (FrameSize * 2) != 0)
{
throw new ArgumentException("Invalid Data Length.");
}
int nbBytes;
short[] input = new short[FrameSize];
byte[] buffer = new byte[200];
byte[] output = new byte[0];
for (int i = 0; i < data.Length / (FrameSize * 2); i++)
{
for (int j = 0; j < input.Length; j++)
{
input[j] = (short)(data[i * FrameSize * 2 + j * 2] + data[i * FrameSize * 2 + j * 2 + 1] * 0x100);
}
nbBytes = Speex.encoder_encode(input, buffer);
Array.Resize<byte>(ref output, output.Length + nbBytes + sizeof(int));
Array.Copy(buffer, 0, output, output.Length - nbBytes, nbBytes);
for (int j = 0; j < sizeof(int); j++)
{
output[output.Length - nbBytes - sizeof(int) + j] = (byte)(nbBytes % 0x100);
nbBytes /= 0x100;
}
}
return output;
}
#endregion
#region Decode
/// <summary>
/// 将编码后的数据进行解码得到原始的音频数据。
/// </summary>
public byte[] Decode(byte[] data)
{
if (this.isDisposed)
{
return null;
}
int nbBytes, index = 0;
byte[] input;
short[] buffer = new short[FrameSize];
byte[] output = new byte[0];
while (index < data.Length)
{
nbBytes = 0;
index += sizeof(int);
for (int i = 1; i <= sizeof(int); i++)
nbBytes = nbBytes * 0x100 + data[index - i];
input = new byte[nbBytes];
Array.Copy(data, index, input, 0, input.Length);
index += input.Length;
Speex.decoder_decode(nbBytes, input, buffer);
Array.Resize<byte>(ref output, output.Length + FrameSize * 2);
for (int i = 0; i < FrameSize; i++)
{
output[output.Length - FrameSize * 2 + i * 2] = (byte)(buffer[i] % 0x100);
output[output.Length - FrameSize * 2 + i * 2 + 1] = (byte)(buffer[i] / 0x100);
}
}
return output;
}
#endregion
#region Pinvoke
[DllImport("Speex.dll", EntryPoint = "encoder_init")]
internal extern static void encoder_init(int quality);
[DllImport("Speex.dll", EntryPoint = "encoder_dispose")]
internal extern static void encoder_dispose();
[DllImport("Speex.dll", EntryPoint = "encoder_encode")]
internal extern static int encoder_encode(short[] data, byte[] output);
[DllImport("Speex.dll", EntryPoint = "decoder_init")]
internal extern static void decoder_init();
[DllImport("Speex.dll", EntryPoint = "decoder_dispose")]
internal extern static void decoder_dispose();
[DllImport("Speex.dll", EntryPoint = "decoder_decode")]
internal extern static void decoder_decode(int nbBytes, byte[] data, short[] output);
#endregion
}
只有四个方法：Initialize、Encode、Decode、Dispose。方法参数的含义也非常明显。
一般音频对话的整个流程是这样的：采集 -> 编码 -> 网络传输 -> 解码 -> 播放。参见：《浅谈网络语音技术》
而该封装的Speex类解决了这个过程中的音频编码和解码的问题。你可以复制该源码到你的项目，并将从http://www.speex.org下载的speex.dll放到运行目录下，就可以正常地使用SPEEX的编解码功能了。
关于Speex更高级的功能，我正在研究中，有兴趣的朋友可以email给我一起探讨。
2012.11.20 我们的研究成果已经全部集成到了OMCS中，其支持回音消除（AEC）、静音检测（VAD）、噪音抑制（DENOISE）、自动增益（AGC）等网络语音技术，有兴趣的可以了解一下。
2014.04.17 Speex dll 点击下载
敬请了解：
ESFramework通信框架 OMCS网络语音视频框架 MFile语音视频录制组件 MCapture语音视频采集组件 StriveEngine轻量级通信引擎 OAUS 自动升级系统
