private bool HasPotentialHumanSpeech(int visualChannel, long pos)
{
Bass.ChannelSetPosition(visualChannel, pos);
const int scaleFactor = _scaleFactorSqr * ushort.MaxValue;
var buffer = new float[2048];
var bufferResult = _isMixerUsed
? BassMix.ChannelGetData(visualChannel, buffer, (int)_maxFft)
: Bass.ChannelGetData(visualChannel, buffer, (int)_maxFft);
var verticalSamplesWithSound = 0;
if (bufferResult > 0)
{
for (var n = 0; n < buffer.Length; n++)
{
// TODO: Don't convert to percentage, just check the threshold raw from the buffer
// TODO: Skip parts of the spectrogram that cannot be human speech
var percentage = (float)Math.Min(1d, Math.Sqrt(buffer[n]) * scaleFactor / ushort.MaxValue);
if (percentage > 0.35)
{
verticalSamplesWithSound++;
if (verticalSamplesWithSound > 4)
{
return(true);
}
}
}
}
return(false);
}
public override int GetData(float[] buffer)
{
var length = default(uint);
switch (buffer.Length)
{
case 128:
length = FFT256;
break;
case 256:
length = FFT512;
break;
case 512:
length = FFT1024;
break;
default:
throw new NotImplementedException();
}
foreach (var channelHandle in this.MixerChannelHandles)
{
return(BassMix.ChannelGetData(channelHandle, buffer, unchecked ((int)length)));
}
return(0);
}
public virtual int GetData(float[] buffer)
{
foreach (var channelHandle in this.GetMixerChannelHandles())
{
//Critical: BassMix.ChannelGetData will trigger an access violation in a random place if called concurrently with different buffer sizes. Yes this took a long time to work out.
lock (ChannelDataSyncRoot)
{
return BassMix.ChannelGetData(channelHandle, buffer, buffer.Length);
}
}
return 0;
}
public virtual int GetData(float[] buffer, int fftSize)
{
var length = default(uint);
switch (fftSize)
{
case BassFFT.FFT256:
length = BassFFT.FFT256_MASK;
break;
case BassFFT.FFT512:
length = BassFFT.FFT512_MASK;
break;
case BassFFT.FFT1024:
length = BassFFT.FFT1024_MASK;
break;
case BassFFT.FFT2048:
length = BassFFT.FFT2048_MASK;
break;
case BassFFT.FFT4096:
length = BassFFT.FFT4096_MASK;
break;
case BassFFT.FFT8192:
length = BassFFT.FFT8192_MASK;
break;
case BassFFT.FFT16384:
length = BassFFT.FFT16384_MASK;
break;
case BassFFT.FFT32768:
length = BassFFT.FFT32768_MASK;
break;
default:
throw new NotImplementedException();
}
foreach (var channelHandle in this.GetMixerChannelHandles())
{
//Critical: BassMix.ChannelGetData will trigger an access violation in a random place if called concurrently with different buffer sizes. Yes this took a long time to work out.
lock (ChannelDataSyncRoot)
{
return(BassMix.ChannelGetData(channelHandle, buffer, unchecked ((int)length)));
}
}
return(0);
}
private int DoWork(AudioVisualizer audioVisualizer, int visualChannel, Work queued, Rectangle clientRectangle)
{
var changes = 0;
var todoWorkList = SplitAndMerge(queued);
var buffer = new float[2048]; // Move back inside loop if behavior is odd
foreach (var work in todoWorkList)
{
try
{
// TODO: Replace all this code with custom BASS_ChannelGetData instead, and paint a heatmap rather than a 3DVoicePrint
_maxFrequencySpectrum = FFTFrequency2Index(audioVisualizer.GetFrequencyRange(), 4096, 44100);
var clientXFrom = audioVisualizer.ByteIndexToClientX(work.From);
var clientXTo = audioVisualizer.ByteIndexToClientX(work.To);
var clientWidth = (int)(clientXTo - clientXFrom);
if (clientWidth <= 0)
{
continue;
}
var bitmapWidth = clientWidth;
var bitmapHeight = clientRectangle.Height;
var bitmap = new DirectBitmap(bitmapWidth, bitmapHeight);
work.Bitmap = bitmap;
changes++;
var clipRectangle = new Rectangle(0, 0, bitmap.Width, bitmap.Height);
if (visualChannel == 0 || clipRectangle.Width <= 1 || clipRectangle.Height <= 1)
{
continue;
}
var bytesPerPixel = (double)(work.To - work.From) / bitmap.Width;
var verticalSamplesPerPixel = (double)_maxFrequencySpectrum / clipRectangle.Height;
var verticalSampleCount = _maxFrequencySpectrum + 1;
const int scaleFactor = _scaleFactorSqr * ushort.MaxValue;
for (var pos = 0; pos < clipRectangle.Width; pos++)
{
Bass.ChannelSetPosition(visualChannel, (long)Math.Floor(work.From + pos * bytesPerPixel));
var bufferResult = _isMixerUsed
? BassMix.ChannelGetData(visualChannel, buffer, (int)_maxFft)
: Bass.ChannelGetData(visualChannel, buffer, (int)_maxFft);
var y1 = 0;
var highest = 0f;
for (var index = 1; index < verticalSampleCount; ++index)
{
if (highest < buffer[index])
{
// By only printing the pixel if we've passed the current pixel's frequencies,
// and keeping the highest frequency of the pixel, we get a better representation.
highest = Math.Max(0, buffer[index]);
}
// TODO: Should not need to divide each time, should be able to increment on each loop.
var currentY = (int)(Math.Round(index / verticalSamplesPerPixel) - 1);
if (currentY > y1)
{
var dbIndex = bitmap.GetStartOffset(pos, y1);
// From near-purple blue (0.70) until red (0.0) in HSV hue wheel.
var percentage = Math.Min(1d, Math.Sqrt(highest) * scaleFactor / ushort.MaxValue);
var cacheIndex = (int)Math.Floor((_cachedColors.Length - 1) * percentage);
var currentColor = _cachedColors[cacheIndex];
bitmap.Bits[dbIndex + 0] = currentColor.B; // B
bitmap.Bits[dbIndex + 1] = currentColor.G; // G
bitmap.Bits[dbIndex + 2] = currentColor.R; // R
bitmap.Bits[dbIndex + 3] = 255; // A
y1 = currentY;
highest = 0;
}
}
if (pos < clipRectangle.Width - 1)
{
for (var y = 0; y < clipRectangle.Height; y++)
{
var dbIndex = bitmap.GetStartOffset(pos + 1, y);
bitmap.Bits[dbIndex + 0] = 0;
bitmap.Bits[dbIndex + 1] = 0;
bitmap.Bits[dbIndex + 2] = 0;
bitmap.Bits[dbIndex + 3] = 255;
}
}
}
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
}
return(changes);
}
public int ChannelGetData(IBassAudioChannel channel, float[] buffer, int length) => BassMix.ChannelGetData(channel.Handle, buffer, length);
