interpolateHRTF

3-D head-related transfer function (HRTF) interpolation

Since R2018b

Syntax

interpolatedHRTF = interpolateHRTF(HRTF,sourcePositions,desiredSourcePositions)

interpolatedHRTF = interpolateHRTF(___,Name,Value)

Description

interpolatedHRTF = interpolateHRTF(HRTF,sourcePositions,desiredSourcePositions) returns the interpolated head-related transfer function (HRTF) at the desired position.

interpolatedHRTF = interpolateHRTF(___,Name,Value) specifies options using one or more Name,Value pair arguments.

Examples

Render 3-D Audio on Headphones

Modify the 3-D audio image of a sound file by filtering it through a head-related transfer function (HRTF). Set the location of the sound source by specifying the desired azimuth and elevation.

load 'ReferenceHRTF.mat' hrtfData sourcePosition

hrtfData = permute(double(hrtfData),[2,3,1]);

sourcePosition = sourcePosition(:,[1,2]);

Calculate the head-related impulse response (HRIR) using the VBAP algorithm at a desired source position. Separate the output, interpolatedIR, into the impulse responses for the left and right ears.

desiredAz = 110;
desiredEl = -45;
desiredPosition = [desiredAz desiredEl];

interpolatedIR  = interpolateHRTF(hrtfData,sourcePosition,desiredPosition, ...
                                  "Algorithm","VBAP");

leftIR = squeeze(interpolatedIR(:,1,:))';
rightIR = squeeze(interpolatedIR(:,2,:))';

Create a dsp.AudioFileReader object to read in a file frame by frame. Create an audioDeviceWriter object to play audio to your sound card frame by frame. Create two dsp.FIRFilter objects and specify the filter coefficients using the head-related transfer function interpolated impulse responses.

fileReader = dsp.AudioFileReader('RockDrums-48-stereo-11secs.mp3');
deviceWriter = audioDeviceWriter('SampleRate',fileReader.SampleRate);

leftFilter = dsp.FIRFilter('Numerator',leftIR);
rightFilter = dsp.FIRFilter('Numerator',rightIR);

In an audio stream loop:

1. Read in a frame of audio data.

2. Feed the stereo audio data through the left and right HRIR filters, respectively.

3. Concatenate the left and right channels and write the audio to your output device.

while ~isDone(fileReader)
    audioIn = fileReader();
    
    leftChannel = leftFilter(audioIn(:,1));
    rightChannel = rightFilter(audioIn(:,2));
    
    deviceWriter([leftChannel,rightChannel]);
end

As a best practice, release your System objects when complete.

release(deviceWriter)
release(fileReader)

Model Moving Source Using HRIR Filtering

Create arrays of head-related impulse responses corresponding to desired source positions. Filter mono input to model a moving source.

Load the ARI HRTF dataset. Cast the hrtfData to type double, and reshape it to the required dimensions: (number of source positions)-by-2-by-(number of HRTF samples). Use the first two columns of the sourcePosition matrix only, which correspond to the azimuth and elevation of the source in degrees.

load 'ReferenceHRTF.mat' hrtfData sourcePosition

hrtfData = permute(double(hrtfData),[2,3,1]);

sourcePosition = sourcePosition(:,[1,2]);

Specify the desired source positions and then calculate the HRTF at these locations using the interpolateHRTF function. Separate the output, interpolatedIR, into the impulse responses for the left and right ears.

desiredAz = [-120;-60;0;60;120;0;-120;120];
desiredEl = [-90;90;45;0;-45;0;45;45];
desiredPosition = [desiredAz desiredEl];

interpolatedIR  = interpolateHRTF(hrtfData,sourcePosition,desiredPosition);

leftIR = squeeze(interpolatedIR(:,1,:));
rightIR = squeeze(interpolatedIR(:,2,:));

Create an audio file sampled at 48 kHz for compatibility with the HRTF dataset.

desiredFs = 48e3;
[audio,fs] = audioread('Counting-16-44p1-mono-15secs.wav');
audio = 0.8*resample(audio,desiredFs,fs);
audiowrite('Counting-16-48-mono-15secs.wav',audio,desiredFs);

Create a dsp.AudioFileReader object to read in a file frame by frame. Create an audioDeviceWriter object to play audio to your sound card frame by frame. Create two dsp.FIRFilter objects with NumeratorSource set to Input port. Setting NumeratorSource to Input port enables you to modify the filter coefficients while streaming.

fileReader = dsp.AudioFileReader('Counting-16-48-mono-15secs.wav');
deviceWriter = audioDeviceWriter('SampleRate',fileReader.SampleRate);

leftFilter = dsp.FIRFilter('NumeratorSource','Input port');
rightFilter = dsp.FIRFilter('NumeratorSource','Input port');

In an audio stream loop:

1. Read in a frame of audio data.

2. Feed the audio data through the left and right HRIR filters.

3. Concatenate the left and right channels and write the audio to your output device. If you have a stereo output hardware, such as headphones, you can hear the source shifting position over time.

4. Modify the desired source position in 2-second intervals by updating the filter coefficients.

durationPerPosition = 2;
samplesPerPosition = durationPerPosition*fileReader.SampleRate;
samplesPerPosition = samplesPerPosition - rem(samplesPerPosition,fileReader.SamplesPerFrame);

sourcePositionIndex = 1;
samplesRead = 0;
while ~isDone(fileReader)
    audioIn = fileReader();
    samplesRead = samplesRead + fileReader.SamplesPerFrame;
    
    leftChannel = leftFilter(audioIn,leftIR(sourcePositionIndex,:));
    rightChannel = rightFilter(audioIn,rightIR(sourcePositionIndex,:));
    
    deviceWriter([leftChannel,rightChannel]);
    
    if mod(samplesRead,samplesPerPosition) == 0
        sourcePositionIndex = sourcePositionIndex + 1;
    end
end

As a best practice, release your System objects when complete.

release(deviceWriter)
release(fileReader)

Input Arguments

HRTF — HRTF values measured at source positions

N-by-2-by-M array

HRTF values measured at the source positions, specified as a N-by-2-by-M array.

• N –– Number of known HRTF pairs

• M –– Number of samples in each known HRTF

If you specify HRTF with real numbers, the function assumes that the input represents an impulse response, and M corresponds to the length of the impulse response. If you specify HRTF with complex numbers, the function assumes that the input represents a transfer function, and M corresponds to the number of bins in the frequency response. The output of the interpolateHRTF function has the same complexity and interpretation as the input.

Data Types: single | double
Complex Number Support: Yes

sourcePositions — Source positions corresponding to measured HRTF values

N-by-2 matrix

Source positions corresponding to measured HRTF values, specified as a N-by-2 matrix. N is the number of known HRTF pairs. The two columns correspond to the azimuth and elevation of the source in degrees, respectively.

Azimuth must be in the range [−180,360]. You can use the −180 to 180 convention or the 0 to 360 convention.

Elevation must be in the range [−90,180]. You can use the −90 to 90 convention or the 0 to 180 convention.

Data Types: single | double

desiredSourcePositions — Desired source positions for HRTF interpolation

P-by-2 matrix

Desired source position for HRTF interpolation, specified as a P-by-2 matrix. P is the number of desired source positions. The columns correspond to the desired azimuth and elevation of the source in degrees, respectively.

Azimuth must be in the range [−180,360]. You can use the −180 to 180 convention or the 0 to 360 convention.

Elevation must be in the range [−90,180]. You can use the −90 to 90 convention or the 0 to 180 convention.

Data Types: single | double

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: 'Algorithm','VBAP'

Algorithm — Interpolation algorithm

'Bilinear' (default) | 'VBAP'

Interpolation algorithm, specified as "Bilinear" or "VBAP".

• Bilinear –– 3-D bilinear interpolation, as specified by [1].

• VBAP –– Vector base amplitude panning interpolation, as specified by [2].

Data Types: char | string

Output Arguments

interpolatedHRTF — Interpolated HRTF

P-by-2-by-M

Interpolated HRTF, returned as a P-by-2-by-M array.

• P –– Number of desired source positions, specified by the number of rows in the desiredSourcePositions input argument.

• M –– Number of samples in each known HRTF, specified by the number of pages in the HRTF input argument.

interpolatedHRTF has the same complexity and interpretation as the input. If you specify the input, HRTF, with real numbers, the function assumes that the input represents an impulse response. If you specify the input with complex numbers, the function assumes that the input represents a transfer function.

Data Types: single | double
Complex Number Support: Yes

References

[1] F.P. Freeland, L.W.P. Biscainho and P.S.R. Diniz, "Interpolation of Head-Related Transfer Functions (HRTFS): A multi-source approach." 2004 12th European Signal Processing Conference. Vienna, 2004, pp. 1761–1764.

[2] Pulkki, Ville. "Virtual Sound Source Positioning Using Vector Based Amplitude Panning." Journal of Audio Engineering Society. Vol. 45. Issue 6, pp. 456–466.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

Introduced in R2018b