Question

I want to have kinect only pick up speech from a user who is directly in front of it.
I don’t want noise or people talking on the right or left to be detected.
Or if the user moves to the right or left of the kinect.
Is this possible?

var audioSource = this.Kinect.AudioSource;
audioSource.BeamAngleMode = BeamAngleMode.Adaptive;
var kinectStream = audioSource.Start();

I have played around with ManualBeamAngle, but do not think that is what I want.

Any help would be appreciated.
cp

Answer 1

You can use the Source angle to find this, they are both 0 when in the middle. See Audio Basics – WPF, and change the code to this:

    double sourceAngle;

    ...

    private void AudioSourceBeamChanged(object sender, BeamAngleChangedEventArgs e)
    {
        beamRotation.Angle = -e.Angle;
        sourceAngle = e.Angle;
        beamAngleText.Text = string.Format(CultureInfo.CurrentCulture, Properties.Resources.BeamAngle, e.Angle.ToString("0", CultureInfo.CurrentCulture));
    }

    private void AudioSourceBeamChanged(object sender, BeamAngleChangedEventArgs e)
    {
        beamRotation.Angle = -e.Angle;
        beamAngleText.Text = string.Format(CultureInfo.CurrentCulture, Properties.Resources.BeamAngle, e.Angle.ToString("0", CultureInfo.CurrentCulture));
    }

    private void AudioReadingThread()
    {
        // Bottom portion of computed energy signal that will be discarded as noise.
        // Only portion of signal above noise floor will be displayed.
        const double EnergyNoiseFloor = 0.2;

        while (this.reading)
        {
            while (sourceAngle == 0) //this is the important part
            {
                int readCount = audioStream.Read(audioBuffer, 0, audioBuffer.Length);

                // Calculate energy corresponding to captured audio in the dispatcher
                // (UI Thread) context, so that rendering code doesn't need to
                // perform additional synchronization.
                Dispatcher.BeginInvoke(
                new Action(
                    () =>
                    {
                        for (int i = 0; i < readCount; i += 2)
                        {
                            // compute the sum of squares of audio samples that will get accumulated
                            // into a single energy value.
                            short audioSample = BitConverter.ToInt16(audioBuffer, i);
                            this.accumulatedSquareSum += audioSample * audioSample;
                            ++this.accumulatedSampleCount;

                            if (this.accumulatedSampleCount < SamplesPerColumn)
                            {
                                continue;
                            }

                            // Each energy value will represent the logarithm of the mean of the
                            // sum of squares of a group of audio samples.
                            double meanSquare = this.accumulatedSquareSum / SamplesPerColumn;
                            double amplitude = Math.Log(meanSquare) / Math.Log(int.MaxValue);

                            // Renormalize signal above noise floor to [0,1] range.
                            this.energy[this.energyIndex] = Math.Max(0, amplitude - EnergyNoiseFloor) / (1 - EnergyNoiseFloor);
                            this.energyIndex = (this.energyIndex + 1) % this.energy.Length;

                            this.accumulatedSquareSum = 0;
                            this.accumulatedSampleCount = 0;
                            ++this.newEnergyAvailable;
                        }
                    }));
            }
        }

The line sourceAngle == 0) does this for us. Hope this helps!