Suppressing Frequencies From FFT

Question

What I am trying to do is to retrieve the frequencies from some song and suppress all the frequencies that do not appear in the human vocal range or in general any range. Here is my suppress function.

    public void SupressAndWrite(Func<FrequencyUnit, bool> func)
    {
        this.WaveManipulated = true;
        while (this.mainWave.WAVFile.NumSamplesRemaining > 0)
        {
            FrequencyUnit[] freqUnits = this.mainWave.NextFrequencyUnits();

            Complex[] compUnits = (from item
                                   in freqUnits
                                   select (func(item)
                                    ? new Complex(item.Frequency, 0) :Complex.Zero))    
                                   .ToArray();

            FourierTransform.FFT(compUnits, FourierTransform.Direction.Backward);

            short[] shorts = (from item
                              in compUnits
                              select (short)item.Real).ToArray();

            foreach (short item in shorts)
            {
                this.ManipulatedFile.AddSample16bit(item);
            }
        }
        this.ManipulatedFile.Close();
    }

Here is my class for my wave.

public sealed class ComplexWave
{
    public readonly WAVFile WAVFile;
    public readonly Int32 SampleSize;

    private FourierTransform.Direction fourierDirection { get; set; }

    private long position;
    /// <param name="file"></param>
    /// <param name="sampleSize in BLOCKS"></param>
    public ComplexWave(WAVFile file, int sampleSize)
    {
        file.NullReferenceExceptionCheck();

        this.WAVFile = file;
        this.SampleSize = sampleSize;

        if (this.SampleSize % 8 != 0)
        {
            if (this.SampleSize % 16 != 0)
            {
                throw new ArgumentException("Sample Size");
            }
        }
        if (!MathTools.IsPowerOf2(sampleSize))
        {
            throw new ArgumentException("Sample Size");
        }
        this.fourierDirection = FourierTransform.Direction.Forward;
    }
    public Complex[] NextSampleFourierTransform()
    {
        short[] newInput = this.GetNextSample();
        Complex[] data = newInput.CopyToComplex();

        if (newInput.Any((x) => x != 0))
        {
            Debug.Write("done");
        }
        FourierTransform.FFT(data, this.fourierDirection);

        return data;
    }
    public FrequencyUnit[] NextFrequencyUnits()
    {
        Complex[] cm = this.NextSampleFourierTransform();
        FrequencyUnit[] freqUn = new FrequencyUnit[(cm.Length / 2)];
        int max = (cm.Length / 2);
        for (int i = 0; i < max; i++)
        {
            freqUn[i] = new FrequencyUnit(cm[i], this.WAVFile.SampleRateHz, i, cm.Length);
        }
        Array.Sort(freqUn);
        return freqUn;
    }
    private short[] GetNextSample()
    {
        short[] retval = new short[this.SampleSize];

        for (int i = 0; i < this.SampleSize; i++)
        {
            if (this.WAVFile.NumSamplesRemaining > 0)
            {
                retval[i] = this.WAVFile.GetNextSampleAs16Bit();
                this.position++;
            }
        }
        return retval;
    }
}

Both FFT forward and FFT backwards work correctly. Could you please tell me what my error is.

Answer 1

Unfortunately, human voice, even when singing, isn't in 'frequency range'. It usually has one main frequency and multitude of harmonics that follow it, depending on the phoneme.

Use this https://play.google.com/store/apps/details?id=radonsoft.net.spectralview&hl=en or some similar app to see what I mean - and then re-define your strategy. Also google 'karaoke' effect.

NEXT:

It's not obvious from your example, but you should scan whole file in windows (google 'fft windowing') to process it whole.