We Keep Coding

Save audio stream float frames as WAV with C#

I am testing an application in C# that receives a live audio stream and then saves it to a WAV file. The audio stream has these characteristics: frequency or sampling rate: 16000, channels: 1, frame Samples Per Channel: 320, play Delay in Ms: 200. The audio frames come as floats, and I am collecting the float frames and storing them into a Memorystream with Binarywriter. After that, I convert the content of the Memorystream into an array, and that array then is converted to a Float array again. With the float array, I start the process to assemble the WAV file.
I have compared the float frames values received with the ones inside the float array that I am using to build the WAV file and are the same. I am having trouble processing the float array to assemble the WAV file. I am not sure if I am doing the data conversion wrong with the ConvertAndWrite() method, or if the WAV header is not well formatted according to the characteristics of the audio stream.
I can see the WAV file being created, but there is no content inside apart from the header I think. Any guidance will be much appreciated. I put together this sample code for you to test what I am doing:
using System;
using System.IO;
using System.Text;
class SaveAudioStreamToWav
{
//Sample as received from stream. Here as a double to avoid altering the sample adding F to each value.
public double[] receivedStreamSample = { 0, -0.003509521, -0.003356934, 0.0002746582, -0.004516602, -0.0027771, -0.0003967285, -0.001739502, 0.004150391, 0.0008544922, 0.002593994, 0.00970459, 0.003631592, 0.001800537, 0.004760742, 0.004272461, -0.002655029, -0.001495361, -0.006835938, -0.004211426, -0.0008239746, 0.001525879, 0.006347656, 0.002532959, -0.002471924, -0.001342773, 0.001159668, 0.0006713867, -0.000793457, 0.001403809, -0.0006713867, -0.0006713867, -0.0007629395, 0.0009460449, -0.003662109, 0.00390625, -0.001312256, -0.001678467, 0.002288818, -0.001831055, -0.00579834, 0.001220703, -0.005096436, -0.003631592, -0.007019043, -0.0001220703, -0.0008850098, -0.0001220703, -0.005371094, 0.004608154, 0.004425049, 0.0027771, 0.005279541, 0.0001525879, 0.0009765625, 0.004150391, -0.002807617, 0.001678467, -0.004577637, -0.002685547, -0.004364014, -0.0008544922, 0.001281738, -0.0009155273, -0.008148193, -0.001983643, 9.155273E-05, 0.0008239746, 0.0004272461, 0.002807617, -0.00289917, 0.002075195, 0.008392334, 0.003479004, 0.005615234, 0.0009460449, 0.002471924, 0.0004272461, -0.006164551, 0.0003967285, -0.0007629395, -0.007476807, -0.002532959, 0.01495361, 0.01382446, 0.002288818, -0.009063721, -0.1129761, -0.05401611, 0.03497314, -0.03027344, 0.08999634, 0.01831055, 0.01037598, 0.03302002, 0.02667236, 0.04309082, -0.01806641, -0.0440979, 0.07125854, 0.00680542, -0.01242065, 0.001983643, -0.03710938, 0.009552002, 0.01013184, 0.002258301, 0.007446289, 0.004486084, -0.009063721, -0.007293701, 0.008239746, -0.0003967285, 0.001556396, 0.001586914, 0.002258301, 0.001281738, 0.001617432, -0.001831055, 0.001556396, -0.001525879, -0.002410889, 0.004516602, 0.000793457, -0.001403809, -0.004882813, -0.0005187988, -0.003540039, -0.004302979, 0.0004272461, 0.004974365, -0.002868652, -0.003875732, -0.0001220703, 0.001617432, 0.002258301, -0.005889893, -0.001068115, 0.003295898, 0.002410889, -0.00201416, 0.001068115, 0.003143311, -0.001464844, 0.000579834, 0.005310059, 0.001434326, 0.001403809, 0.001312256, -0.001617432, 0.0009460449, -0.0009765625, -0.0007324219, -0.001617432, -0.004730225, 0.001373291, -0.001586914, 0.0005187988, 0.001556396, -0.001647949, 0.0008544922, 0.001739502, 0.0027771, 0.001831055, 3.051758E-05, -0.04672241, 0.02276611, 0.02529907, -0.005249023, -0.02285767, -0.0378418, -0.1454468, 0.04385376, -0.04058838, -0.005249023, -3.051758E-05, -0.02166748, -0.006378174, -0.002380371, -0.0368042, 0.04330444, -0.008453369, 0.0300293, -0.01651001, -0.005554199, -0.01828003, 0.008972168, -0.01571655, -0.01202393, 0.01141357, -0.003997803, 0.004119873, -0.002532959, 0.004333496, -0.001495361, -0.001281738, -0.003692627, -0.001647949, -0.001861572, 0.000793457, -0.0003662109, -0.002532959, -0.001342773, 0.0003051758, 0.002075195, 0.002349854, 0.001464844, 0.001678467, -0.0008850098, -0.0001525879, 0.003723145, -0.0009155273, 0.002807617, -0.005157471, -0.001617432, 0.002471924, 0.002166748, -0.0003356934, 0.000213623, -0.000793457, -0.0008544922, -0.00100708, 0.000213623, 0.001037598, -0.003448486, 0.0009460449, -0.0006103516, -0.002655029, -0.009735107, -0.01101685, 0.01937866, 0.00994873, -0.02600098, 0.04592896, 0.1063843, 0.002441406, -0.0100708, 0.002990723, -0.01235962, -0.003448486, 0.01089478, -0.01480103, -0.02902222, 0.02990723, -0.01376343, 0.01275635, -0.008666992, 0.006469727, -0.009857178, 0.002655029, -0.0004882813, 0.003814697, 0.004943848, -0.002990723, -0.0003051758, -0.001678467, 0.003265381, 0.0009460449, -9.155273E-05, -0.001403809, 0.001739502, -0.002685547, -0.0009460449, -0.001281738, 0.0009765625, 0.001312256, 0.002288818, -0.0002746582, -0.001098633, -0.002319336, -0.000793457, 0.001464844, 0.001281738, -0.002319336, 6.103516E-05, 0.0003967285, -0.002532959, 0.0002441406, 0.001861572, 0.0009765625 };
public float[] floatsArray;
public FileStream fileStream;
static void Main(string[] args)
{
var saveAudioStreamToWav = new SaveAudioStreamToWav();
saveAudioStreamToWav.ConvertDoubleToFloat();
saveAudioStreamToWav.CreateEmpty(saveAudioStreamToWav.SetNameAndPath());
saveAudioStreamToWav.ConvertAndWrite();
saveAudioStreamToWav.WriteHeader();
}
public void ConvertDoubleToFloat()
{
floatsArray = new float[receivedStreamSample.Length];
floatsArray = Array.ConvertAll(receivedStreamSample, x => (float)x);
}
public string SetNameAndPath()
{
//Setting the name of the file
string timeStamp = DateTime.Now.ToString("yyyyMMddHHmmssfff");
string filename = "/TestSavingStreamToWav_" + timeStamp + ".wav";
string path = Directory.GetCurrentDirectory();
string filepath = path + filename;
Console.WriteLine(filepath);
return filepath;
}
public void CreateEmpty(string filepath)
{
const int HEADER_SIZE = 44;
fileStream = new FileStream(filepath, FileMode.CreateNew, FileAccess.ReadWrite);
byte emptyByte = new byte();
for (int i = 0; i < HEADER_SIZE; i++) //preparing an empty space for the header
{
fileStream.WriteByte(emptyByte);
}
}
public void ConvertAndWrite()
{
Int16[] intData = new Int16[floatsArray.Length];
Byte[] bytesData = new Byte[floatsArray.Length * 2]; // bytesData array is twice the size of floatsArray array because a float converted in Int16 is 2 bytes.
const float rescaleFactor = 32767; //to convert float to Int16
for (var i = 0; i < floatsArray.Length; i++)
{
intData[i] = (short)(floatsArray[i] * rescaleFactor);
var byteArr = new Byte[2];
byteArr = BitConverter.GetBytes(intData[i]);
byteArr.CopyTo(bytesData, i * 2);
}
fileStream.Write(bytesData, 0, bytesData.Length);
}
public void WriteHeader()
{
int hz = 16000; //frequency or sampling rate
int headerSize = 44; //default for uncompressed wav
fileStream.Seek(0, SeekOrigin.Begin);
Byte[] riff = System.Text.Encoding.UTF8.GetBytes("RIFF"); //RIFF marker. Marks the file as a riff file. Characters are each 1 byte long.
fileStream.Write(riff, 0, 4);
Byte[] chunkSize = BitConverter.GetBytes(fileStream.Length - 8); //file-size (equals file-size - 8). Size of the overall file - 8 bytes, in bytes (32-bit integer). Typically, you'd fill this in after creation.
fileStream.Write(chunkSize, 0, 4);
Byte[] wave = System.Text.Encoding.UTF8.GetBytes("WAVE"); //File Type Header. For our purposes, it always equals "WAVE".
fileStream.Write(wave, 0, 4);
Byte[] fmt = System.Text.Encoding.UTF8.GetBytes("fmt "); //Mark the format section. Format chunk marker. Includes trailing null.
fileStream.Write(fmt, 0, 4);
Byte[] subChunk1 = BitConverter.GetBytes(16); //Length of format data. Always 16.
fileStream.Write(subChunk1, 0, 4);
UInt16 two = 2;
UInt16 one = 1;
Byte[] audioFormat = BitConverter.GetBytes(one); //Type of format (1 is PCM, other number means compression) . 2 byte integer. Wave type PCM
fileStream.Write(audioFormat, 0, 2);
Byte[] numChannels = BitConverter.GetBytes(one); //Number of Channels - 2 byte integer
fileStream.Write(numChannels, 0, 2);
Byte[] sampleRate = BitConverter.GetBytes(hz); //Sample Rate - 32 byte integer. Sample Rate = Number of Samples per second, or Hertz.
fileStream.Write(sampleRate, 0, 4);
Byte[] byteRate = BitConverter.GetBytes(hz * 2 * 1);// sampleRate * bytesPerSample * number of channels, here 16000*2*1.
fileStream.Write(byteRate, 0, 4);
UInt16 blockAlign = (ushort)(1 * 2); //channels * bytesPerSample, here 1 * 2 // Bytes Per Sample: 1=8 bit Mono, 2 = 8 bit Stereo or 16 bit Mono, 4 = 16 bit Stereo
fileStream.Write(BitConverter.GetBytes(blockAlign), 0, 2);
UInt16 sixteen = 16;
Byte[] bitsPerSample = BitConverter.GetBytes(sixteen); //Bits per sample (BitsPerSample * Channels) ?? should be 8???
fileStream.Write(bitsPerSample, 0, 2);
Byte[] dataString = System.Text.Encoding.UTF8.GetBytes("data"); //"data" chunk header. Marks the beginning of the data section.
fileStream.Write(dataString, 0, 4);
Byte[] subChunk2 = BitConverter.GetBytes(fileStream.Length - headerSize); //Size of the data section. data-size (equals file-size - 44). or NumSamples * NumChannels * bytesPerSample ??
fileStream.Write(subChunk2, 0, 4);
fileStream.Close();
}
}//end of class

I have updated your code into an extension method.
The idea was so you could append you data to a stream, like a file stream or memory stream, obviously this won't work for non seekable streams. So you could probably add error checking and validation.
I think I got the header right after looking at the specs, it seems to play at least. Note this is not really cross platform because of the endianness.
I'm not really sure what the rescaleFactor however I'll have to trust you there.
However, you should be able to modify this to accept data in different formats.
Lastly, I am updating the header at the end of the append, you could probably do this separately, i.e keep adding to the stream and then update it once when finished, add pepper and salt to taste.
Usage
using (var stream = new FileStream(GetFileName(), FileMode.OpenOrCreate, FileAccess.ReadWrite))
{
stream.AppendWaveData(receivedStreamSample);
}
Extension
public static class BinaryWriterExtensions
{
private const int HeaderSize = 44;
private const int Hz = 16000; //frequency or sampling rate
private const float RescaleFactor = 32767; //to convert float to Int16
public static void AppendWaveData<T>(this T stream, float[] buffer)
where T : Stream
{
if (stream.Length > HeaderSize)
{
stream.Seek(0, SeekOrigin.End);
}
else
{
stream.SetLength(HeaderSize);
stream.Position = HeaderSize;
}
// rescale
var floats = Array.ConvertAll(buffer, x => (short)(x * RescaleFactor));
// Copy to bytes
var result = new byte[floats.Length * sizeof(short)];
Buffer.BlockCopy(floats, 0, result, 0, result.Length);
// write to stream
stream.Write(result, 0, result.Length);
// Update Header
UpdateHeader(stream);
}
public static void UpdateHeader(Stream stream)
{
var writer = new BinaryWriter(stream);
writer.Seek(0, SeekOrigin.Begin);
writer.Write(Encoding.ASCII.GetBytes("RIFF")); //RIFF marker. Marks the file as a riff file. Characters are each 1 byte long.
writer.Write((int)(writer.BaseStream.Length - 8)); //file-size (equals file-size - 8). Size of the overall file - 8 bytes, in bytes (32-bit integer). Typically, you'd fill this in after creation.
writer.Write(Encoding.ASCII.GetBytes("WAVE")); //File Type Header. For our purposes, it always equals "WAVE".
writer.Write(Encoding.ASCII.GetBytes("fmt ")); //Mark the format section. Format chunk marker. Includes trailing null.
writer.Write(16); //Length of format data. Always 16.
writer.Write((short)1); //Type of format (1 is PCM, other number means compression) . 2 byte integer. Wave type PCM
writer.Write((short)2); //Number of Channels - 2 byte integer
writer.Write(Hz); //Sample Rate - 32 byte integer. Sample Rate = Number of Samples per second, or Hertz.
writer.Write(Hz * 2 * 1); // sampleRate * bytesPerSample * number of channels, here 16000*2*1.
writer.Write((short)(1 * 2)); //channels * bytesPerSample, here 1 * 2 // Bytes Per Sample: 1=8 bit Mono, 2 = 8 bit Stereo or 16 bit Mono, 4 = 16 bit Stereo
writer.Write((short)16); //Bits per sample (BitsPerSample * Channels) ?? should be 8???
writer.Write(Encoding.ASCII.GetBytes("data")); //"data" chunk header. Marks the beginning of the data section.
writer.Write((int)(writer.BaseStream.Length - HeaderSize)); //Size of the data section. data-size (equals file-size - 44). or NumSamples * NumChannels * bytesPerSample ??
}
} //end of class

change
writer.Write((short)2); into writer.Write((short)1);
and the generated file(.wav) will be plays well both on Windows(test against on windows 7) and android device.
otherwise,the Windows Media Player will says:having trouble playing files;android will plays with a quickly speed than expected.

how to get PCM data from stereo channel mp3 using Naudio in C#

I am new to Naudio and using it to get PCM data from Mp3 files, this is my code to take PCM from mono-channel file, but don't know how to do it with stereo channel file
code:
Mp3FileReader file = new Mp3FileReader(op.FileName);
int _Bytes = (int)file.Length;
byte[] Buffer = new byte[_Bytes];
file.Read(Buffer, 0, (int)_Bytes);
for (int i = 0; i < Buffer.Length - 2; i += 2)
{
byte[] Sample_Byte = new byte[2];
Sample_Byte[0] = Buffer[i + 1];
Sample_Byte[1] = Buffer[i + 2];
Int16 _ConvertedSample = BitConverter.ToInt16(Sample_Byte, 0);
}
How can I get PCM from stereo channel Mp3 file?

In a stereo file, the samples are interleaved: one left channel sample followed by one right channel etc. So in your loop you could go through four bytes at a time to read out the samples.
Also there are some bugs in your code. You should use return value of Read, not the size of the buffer, and you have an off by one error in the code to access the samples. Also, no need to copy into a temporary buffer.
Something like this should work for you:
var file = new Mp3FileReader(fileName);
int _Bytes = (int)file.Length;
byte[] Buffer = new byte[_Bytes];
int read = file.Read(Buffer, 0, (int)_Bytes);
for (int i = 0; i < read; i += 4)
{
Int16 leftSample = BitConverter.ToInt16(Buffer, i);
Int16 rightSample = BitConverter.ToInt16(Buffer, i + 2);
}

Writing the wav header for multiple channels

Following the wave file standard to record audio data with Kinect, I wrote a WriteWavHeader() method which works perfect for 1 channel (PCM, Sample rate: 16 kHz, Bits/sample: 16). The problem is that when I try to record in stereo (nChannels = 2) it seems like the speed of the track doubles! Any ideas what I am doing wrong?
static void WriteWavHeader(Stream stream, int dataLength)
{
using (var memStream = new MemoryStream(64))
{
int cbFormat = 16; //sizeof(WAVEFORMATEX)
WAVEFORMATEX format = new WAVEFORMATEX()
{
//Subchunk1Size == 16 for PCM.
wFormatTag = 1,
//NumChannels : Mono = 1, Stereo = 2, etc.
nChannels = 2,
//SampleRate 8000, 44100
nSamplesPerSec = 16000,
//ByteRate == SampleRate * NumChannels * BitsPerSample/8
nAvgBytesPerSec = 32000*2,
//BlockAlign == NumChannels * BitsPerSample/8
nBlockAlign = 2*2,
//BitsPerSample 8 bits = 8, 16 bits = 16, etc.
wBitsPerSample = 16
};
using (var binarywriter = new BinaryWriter(memStream))
{
//RIFF header
WriteString(memStream, "RIFF");
binarywriter.Write(dataLength+8 + cbFormat+8 + 4);
WriteString(memStream, "WAVE");
WriteString(memStream, "fmt ");
binarywriter.Write(cbFormat);
//WAVEFORMATEX
binarywriter.Write(format.wFormatTag);
binarywriter.Write(format.nChannels);
binarywriter.Write(format.nSamplesPerSec);
binarywriter.Write(format.nAvgBytesPerSec);
binarywriter.Write(format.nBlockAlign);
binarywriter.Write(format.wBitsPerSample);
//data header
WriteString(memStream, "data");
binarywriter.Write(dataLength);
memStream.WriteTo(stream);
}
}
}
And the recording method:
public void RecordAudio()
{
//Subchunk2Size == NumSamples * NumChannels * BitsPerSample/8
int recordingLength = 5 *16000* 2 *2;
byte[] buffer = new byte[1024];
using (FileStream _fileStream = new FileStream("c:\\kinectAudio.wav", FileMode.Create))
{
WriteWavHeader(_fileStream, recordingLength);
int count, totalCount = 0;
//Start capturing audio
using (Stream audioStream = this.sensor.AudioSource.Start())
{
while ((count = audioStream.Read(buffer, 0, buffer.Length)) > 0 && totalCount < recordingLength)
{
_fileStream.Write(buffer, 0, count);
totalCount += count;
}
}
//write the real wav header
long prePosition = _fileStream.Position;
_fileStream.Seek(0, SeekOrigin.Begin);
WriteWavHeader(_fileStream, totalCount);
_fileStream.Seek(prePosition, SeekOrigin.Begin);
_fileStream.Flush();
}
}

c# Pitch shift of wave files

I'm currently trying to do pitch shifting of a wave file using this algorithm
https://sites.google.com/site/mikescoderama/pitch-shifting
Here my code which use the above implementation, but with no luck. The outputted wave file seems to be corrupted or not valid.
The code is quite simple, except for the pitch shift algorithm :)
It load a wave file, it reads the wave file data and put it in a
byte[] array.
Then it "normalize" bytes data into -1.0f to 1.0f format (as
requested by the creator of the pitch shift algorithm).
It applies the pitch shift algorithm and then convert back the
normalized data into a bytes[] array.
Finally saves a wave file with the same header of the original wave
file and the pitch shifted data.
Am I missing something?
static void Main(string[] args)
{
// Read the wave file data bytes
byte[] waveheader = null;
byte[] wavedata = null;
using (BinaryReader reader = new BinaryReader(File.OpenRead("sound.wav")))
{
// Read first 44 bytes (header);
waveheader= reader.ReadBytes(44);
// Read data
wavedata = reader.ReadBytes((int)reader.BaseStream.Length - 44);
}
short nChannels = BitConverter.ToInt16(waveheader, 22);
int sampleRate = BitConverter.ToInt32(waveheader, 24);
short bitRate = BitConverter.ToInt16(waveheader, 34);
// Normalized data store. Store values in the format -1.0 to 1.0
float[] in_data = new float[wavedata.Length / 2];
// Normalize wave data into -1.0 to 1.0 values
using(BinaryReader reader = new BinaryReader(new MemoryStream(wavedata)))
{
for (int i = 0; i < in_data.Length; i++)
{
if(bitRate == 16)
in_data[i] = reader.ReadInt16() / 32768f;
if (bitRate == 8)
in_data[i] = (reader.ReadByte() - 128) / 128f;
}
}
//PitchShifter.PitchShift(1f, in_data.Length, (long)1024, (long)32, sampleRate, in_data);
// Backup wave data
byte[] copydata = new byte[wavedata.Length];
Array.Copy(wavedata, copydata, wavedata.Length);
// Revert data to byte format
Array.Clear(wavedata, 0, wavedata.Length);
using (BinaryWriter writer = new BinaryWriter(new MemoryStream(wavedata)))
{
for (int i = 0; i < in_data.Length; i++)
{
if(bitRate == 16)
writer.Write((short)(in_data[i] * 32768f));
if (bitRate == 8)
writer.Write((byte)((in_data[i] * 128f) + 128));
}
}
// Compare new wavedata with copydata
if (wavedata.SequenceEqual(copydata))
{
Console.WriteLine("Data has no changes");
}
else
{
Console.WriteLine("Data has changed!");
}
// Save modified wavedata
string targetFilePath = "sound_low.wav";
if (File.Exists(targetFilePath))
File.Delete(targetFilePath);
using (BinaryWriter writer = new BinaryWriter(File.OpenWrite(targetFilePath)))
{
writer.Write(waveheader);
writer.Write(wavedata);
}
Console.ReadLine();
}

The algorithm here works fine
https://sites.google.com/site/mikescoderama/pitch-shifting
My mistake was on how i was reading the wave header and wave data. I post here the fully working code
WARNING: this code works only for PCM 16 bit (stereo/mono) waves. Can be easily adapted to works with PCM 8 bit.
static void Main(string[] args)
{
// Read header, data and channels as separated data
// Normalized data stores. Store values in the format -1.0 to 1.0
byte[] waveheader = null;
byte[] wavedata = null;
int sampleRate = 0;
float[] in_data_l = null;
float[] in_data_r = null;
GetWaveData("sound.wav", out waveheader, out wavedata, out sampleRate, out in_data_l, out in_data_r);
//
// Apply Pitch Shifting
//
if(in_data_l != null)
PitchShifter.PitchShift(2f, in_data_l.Length, (long)1024, (long)10, sampleRate, in_data_l);
if(in_data_r != null)
PitchShifter.PitchShift(2f, in_data_r.Length, (long)1024, (long)10, sampleRate, in_data_r);
//
// Time to save the processed data
//
// Backup wave data
byte[] copydata = new byte[wavedata.Length];
Array.Copy(wavedata, copydata, wavedata.Length);
GetWaveData(in_data_l, in_data_r, ref wavedata);
//
// Check if data actually changed
//
bool noChanges = true;
for (int i = 0; i < wavedata.Length; i++)
{
if (wavedata[i] != copydata[i])
{
noChanges = false;
Console.WriteLine("Data has changed!");
break;
}
}
if(noChanges)
Console.WriteLine("Data has no changes");
// Save modified wavedata
string targetFilePath = "sound_low.wav";
if (File.Exists(targetFilePath))
File.Delete(targetFilePath);
using (BinaryWriter writer = new BinaryWriter(File.OpenWrite(targetFilePath)))
{
writer.Write(waveheader);
writer.Write(wavedata);
}
Console.ReadLine();
}
// Returns left and right float arrays. 'right' will be null if sound is mono.
public static void GetWaveData(string filename, out byte[] header, out byte[] data, out int sampleRate, out float[] left, out float[] right)
{
byte[] wav = File.ReadAllBytes(filename);
// Determine if mono or stereo
int channels = wav[22]; // Forget byte 23 as 99.999% of WAVs are 1 or 2 channels
// Get sample rate
sampleRate = BitConverter.ToInt32(wav, 24);
int pos = 12;
// Keep iterating until we find the data chunk (i.e. 64 61 74 61 ...... (i.e. 100 97 116 97 in decimal))
while(!(wav[pos]==100 && wav[pos+1]==97 && wav[pos+2]==116 && wav[pos+3]==97)) {
pos += 4;
int chunkSize = wav[pos] + wav[pos + 1] * 256 + wav[pos + 2] * 65536 + wav[pos + 3] * 16777216;
pos += 4 + chunkSize;
}
pos += 4;
int subchunk2Size = BitConverter.ToInt32(wav, pos);
pos += 4;
// Pos is now positioned to start of actual sound data.
int samples = subchunk2Size / 2; // 2 bytes per sample (16 bit sound mono)
if (channels == 2)
samples /= 2; // 4 bytes per sample (16 bit stereo)
// Allocate memory (right will be null if only mono sound)
left = new float[samples];
if (channels == 2)
right = new float[samples];
else
right = null;
header = new byte[pos];
Array.Copy(wav, header, pos);
data = new byte[subchunk2Size];
Array.Copy(wav, pos, data, 0, subchunk2Size);
// Write to float array/s:
int i=0;
while (pos < subchunk2Size)
{
left[i] = BytesToNormalized_16(wav[pos], wav[pos + 1]);
pos += 2;
if (channels == 2)
{
right[i] = BytesToNormalized_16(wav[pos], wav[pos + 1]);
pos += 2;
}
i++;
}
}
// Return byte data from left and right float data. Ignore right when sound is mono
public static void GetWaveData(float[] left, float[] right, ref byte[] data)
{
// Calculate k
// This value will be used to convert float to Int16
// We are not using Int16.Max to avoid peaks due to overflow conversions
float k = (float)Int16.MaxValue / left.Select(x => Math.Abs(x)).Max();
// Revert data to byte format
Array.Clear(data, 0, data.Length);
int dataLenght = left.Length;
int byteId = -1;
using (BinaryWriter writer = new BinaryWriter(new MemoryStream(data)))
{
for (int i = 0; i < dataLenght; i++)
{
byte byte1 = 0;
byte byte2 = 0;
byteId++;
NormalizedToBytes_16(left[i], k, out byte1, out byte2);
writer.Write(byte1);
writer.Write(byte2);
if (right != null)
{
byteId++;
NormalizedToBytes_16(right[i], k, out byte1, out byte2);
writer.Write(byte1);
writer.Write(byte2);
}
}
}
}
// Convert two bytes to one double in the range -1 to 1
static float BytesToNormalized_16(byte firstByte, byte secondByte)
{
// convert two bytes to one short (little endian)
short s = (short)((secondByte << 8) | firstByte);
// convert to range from -1 to (just below) 1
return s / 32678f;
}
// Convert a float value into two bytes (use k as conversion value and not Int16.MaxValue to avoid peaks)
static void NormalizedToBytes_16(float value, float k, out byte firstByte, out byte secondByte)
{
short s = (short)(value * k);
firstByte = (byte)(s & 0x00FF);
secondByte = (byte)(s >> 8);
}

sorry to revive this but I tried that pitchshifter class and, while it works, I get crackles in the audio while pitching down(0.5f). You work out a way around that?

Windows Phone Encoding and Decoding audio using NSpeex. Having issue with decoding?

I am trying to encode a recorded audio using Nspeex and then transfer it over internet and decode on the other end. I am doing all this in Windows Phone 7/8. To encode and decode I am using following code. But while decoding I am not getting the result back correctly which I can play again. Can anyone provide me with encoding and decoding code which runs on WP7/8 recorded audio:
private static Microphone mic = Microphone.Default;
private static byte[] EncodeSpeech(byte[] buf, int len)
{
BandMode mode = GetBandMode(mic.SampleRate);
SpeexEncoder encoder = new SpeexEncoder(mode);
// set encoding quality to lowest (which will generate the smallest size in the fastest time)
encoder.Quality = 1;
int inDataSize = len / 2;
// convert to short array
short[] data = new short[inDataSize];
int sampleIndex = 0;
for (int index = 0; index < len; index += 2, sampleIndex++)
{
data[sampleIndex] = BitConverter.ToInt16(buf, index);
}
// note: the number of samples per frame must be a multiple of encoder.FrameSize
inDataSize = inDataSize - inDataSize % encoder.FrameSize;
var encodedData = new byte[len];
int encodedBytes = encoder.Encode(data, 0, inDataSize, encodedData, 0, len);
if (encodedBytes != 0)
{
// each chunk is laid out as follows:
// | 4-byte total chunk size | 4-byte encoded buffer size | <encoded-bytes> |
byte[] inDataSizeBuf = BitConverter.GetBytes(inDataSize);
byte[] sizeBuf = BitConverter.GetBytes(encodedBytes + inDataSizeBuf.Length);
byte[] returnBuf = new byte[encodedBytes + sizeBuf.Length + inDataSizeBuf.Length];
sizeBuf.CopyTo(returnBuf, 0);
inDataSizeBuf.CopyTo(returnBuf, sizeBuf.Length);
Array.Copy(encodedData, 0, returnBuf, sizeBuf.Length + inDataSizeBuf.Length, encodedBytes);
return returnBuf;
}
else
return buf;
}
private byte[] DecodeSpeech(byte[] buf)
{
BandMode mode = GetBandMode(mic.SampleRate);
SpeexDecoder decoder = new SpeexDecoder(mode);
byte[] inDataSizeBuf = new byte[4];
byte[] sizeBuf = new byte[4];
byte[] encodedBuf = new byte[buf.Length - 8];
Array.Copy(buf, 0, sizeBuf, 0, 4);
Array.Copy(buf, 4, inDataSizeBuf, 0, 4);
Array.Copy(buf, 8, encodedBuf, 0, buf.Length - 8);
int inDataSize = BitConverter.ToInt32(inDataSizeBuf, 0);
int size = BitConverter.ToInt32(sizeBuf, 0);
short[] decodedBuf = new short[inDataSize];
int decodedSize = decoder.Decode(encodedBuf, 0, encodedBuf.Length, decodedBuf, 0, false);
byte[] returnBuf = new byte[inDataSize * 2];
for (int index = 0; index < decodedBuf.Length; index++)
{
byte[] temp = BitConverter.GetBytes(decodedBuf[index]);
Array.Copy(temp, 0, returnBuf, index * 2, 2);
}
return returnBuf;
}
private static BandMode GetBandMode(int sampleRate)
{
if (sampleRate <= 8000)
return BandMode.Narrow;
if (sampleRate <= 16000)
return BandMode.Wide;
return BandMode.UltraWide;
}

I think your problem may be that you are newing up a new SpeexEncoder every time you want to encode audio. You should try making that a member for your class and re-use it.
I looked at the code for Nspeex I noticed that SpeexEncoder uses NbEncoder for the narrow band. In that class it looks like it keeps a history of some previous audio data in order perform the encoding. This should mean that the output for different instances of encoders would not go together.
private static Microphone mic = Microphone.Default;
private static SpeexEncoder encoder = CreateEncoder();
private static SpeexEncoder CreateEncoder()
{
BandMode mode = GetBandMode(mic.SampleRate);
SpeexEncoder encoder = new SpeexEncoder(mode);
// set encoding quality to lowest (which will generate the smallest size in the fastest time)
encoder.Quality = 1;
return encoder;
}
private static byte[] EncodeSpeech(byte[] buf, int len)
{
int inDataSize = len / 2;
...