% SUpDEq - Spatial Upsampling by Directional Equalization
script supdeq_demo_ALFE
This script shows how the ALFE (Adaptive Low Frequency Extention)
algorithm can be applied in combination with SUpDEq.
Reference:
C. Pörschmann and J. M. Arend, "Obtaining Dense HRTF Sets from Sparse
Measurements in Reverberant Environments," in Proceedings of the
AES International Conference on Immersive and Interactive Audio,
York, UK, 2019, pp. 1?10.
(C) 2019 by JMA, Johannes M. Arend
CP, Christoph Pörschmann
Technische Hochschule Köln
University of Applied Sciences
Institute of Communications Engineering
Department of Acoustics and Audio Signal Processing
% (1) - Load reference HRTF dataset (stored as SH-coefficients in spherical harmonics domain)
See publication for more info about the measurement setup of the reference HRTF dataset.
The reference dataset is only needed for comparison with the final de-equalized dataset...0001 %% SUpDEq - Spatial Upsampling by Directional Equalization 0002 % 0003 % script supdeq_demo_ALFE 0004 % 0005 % This script shows how the ALFE (Adaptive Low Frequency Extention) 0006 % algorithm can be applied in combination with SUpDEq. 0007 % 0008 % Reference: 0009 % C. Pörschmann and J. M. Arend, "Obtaining Dense HRTF Sets from Sparse 0010 % Measurements in Reverberant Environments," in Proceedings of the 0011 % AES International Conference on Immersive and Interactive Audio, 0012 % York, UK, 2019, pp. 1?10. 0013 % 0014 % (C) 2019 by JMA, Johannes M. Arend 0015 % CP, Christoph Pörschmann 0016 % Technische Hochschule Köln 0017 % University of Applied Sciences 0018 % Institute of Communications Engineering 0019 % Department of Acoustics and Audio Signal Processing 0020 % 0021 %% (1) - Load reference HRTF dataset (stored as SH-coefficients in spherical harmonics domain) 0022 %See publication for more info about the measurement setup of the reference HRTF dataset. 0023 %The reference dataset is only needed for comparison with the final de-equalized dataset... 0024 referenceHRTFdataset = importdata('RAR_HRIR_1m_Genelec_2702SFD.mat'); 0025 0026 %% (2) - Load sparse HRTF dataset (stored as FFT-bins in Fourier domain) 0027 %See publication for more info about the measurement setup of the sparse HRTF dataset. 0028 load('ZW8_3_HRIR_1m_Genelec_sp_L86.mat'); 0029 0030 %% (3) - Get equalization dataset (SH-coefficients) 0031 %The eqDataset describes the sound pressure distribution on a sphere 0032 %Use defaults: N = 35, earDistance = 0.165m, NFFT = 512, fs = 48000; 0033 NFFT=(length(sparseHRTFdataset.HRTF_L(1,:))-1)*2; 0034 eqDataset = supdeq_getEqDataset(35,0.165,NFFT,48000); 0035 0036 %% (4) - Perform equalization 0037 %Here, the sparse HRTF dataset is equalized with the eqDataset. The 0038 %equalized HRTF are transformed to the spherical harmonics domain again with 0039 %the maximal order N which is possible with the sparse sampling grid. 0040 %N and the sparse sampling grid are part of the sparseHRTFdataset struct 0041 sparseSamplingGrid = sparseHRTFdataset.samplingGrid; 0042 Nsparse = sparseHRTFdataset.Nmax; 0043 0044 eqHRTFdataset = supdeq_eq(sparseHRTFdataset,eqDataset,Nsparse,sparseSamplingGrid); 0045 0046 %% (4a) - Perform ALFE - If desired 0047 %In this section the lower frequency components below e certain frequency 0048 %xover are replaced by a frequecy-shaped dirac. The ALFE algorithm is done 0049 %on the equalized dataset in order to benefit from the constant frequency 0050 %response towards low frequencies. 0051 xover = [500 750]; 0052 fs = 48000; 0053 [eqHRTFdataset_ALFE] = supdeq_alfe(eqHRTFdataset,sparseHRTFdataset.samplingGrid,xover,fs); 0054 eqHRTFdataset=eqHRTFdataset_ALFE; 0055 0056 %% (5) - Perform de-equalization 0057 %Here, the sparse equalized HRTF dataset is de-equalized with the 0058 %deqDataset. This is done on a dense spatial sampling grid. The results is a 0059 %dense HRTF/HRIR dataset. In this case, deqDataset is the same as the 0060 %eqDataset... 0061 0062 %First, define dense spatial sampling grid. Here, we use the lebedev grid 0063 %with 2702 points again (same as the reference HRIR dataset). 0064 %The highest stable grid order here is N = 44. However, we use N = 35 for the 0065 %spherical Fourier transform of the de-equalized HRTFs. 0066 denseSamplingGrid = supdeq_lebedev(2702); 0067 Ndense = 35; 0068 0069 %Perform de-equalization. Apply head and tail window (8 and 32 samples 0070 %respectively) to de-equalized HRIRs/HRTFs. 0071 [denseHRTFdataset, denseHRIRdataset, denseHRTFdataset_sh] = supdeq_deq(eqHRTFdataset, eqDataset, Ndense, denseSamplingGrid,[8,32]); 0072 0073 %% (6) - Optional: Save as SOFA object 0074 %Use defaults: fs = 48000, earDistance = 0.165m, sourceDistance = 3.0m 0075 %denseHRIRdataset_SOFA = supdeq_writeSOFAobj(denseHRIRdataset.HRIR_L,denseHRIRdataset.HRIR_R,denseSamplingGrid); 0076 0077 %% (6a) - Optional: Save SH coefficients 0078 save ('denseHRTFdataset_JBL_L2702.mat', 'denseHRTFdataset_sh') 0079 0080 %% (7) - Optional: Plot HRIRs 0081 %Get HRIRs from reference dataset and de-equalized dense dataset. In this 0082 %example, we chose a lateral source, because most differences between the 0083 %reference and the de-equalized HRIRs occure at the contralateral ear. 0084 azPlot = 180; 0085 elPlot = 90; 0086 0087 [hrir_ref(:,1),hrir_ref(:,2)] = supdeq_getArbHRIR(referenceHRTFdataset,[azPlot,elPlot],'DEG',2,'ak'); 0088 [hrir_deq(:,1),hrir_deq(:,2)] = supdeq_getArbHRIR(denseHRTFdataset_sh,[azPlot,elPlot],'DEG',2,'ak'); 0089 0090 supdeq_plotIR(hrir_ref(:,2),hrir_deq(:,2),[],[],1); 0091 0092 %% (8) - Optional: Listen to the results 0093 %Here, you can listen to a short drums sequence, spatialized with a 0094 %reference HRIR and a de-equalized HRIR 0095 0096 %Load drums test signal 0097 [testSignal,fsTestSignal] = audioread('drums.wav'); 0098 0099 %Define az and el for playback 0100 azPlayback = 270; 0101 elPlayback = 90; 0102 0103 %Convolve with reference HRIR and play back 0104 supdeq_listen(referenceHRTFdataset,testSignal,[azPlayback,elPlayback]); 0105 %Wait for length of test-signal 0106 pause(length(testSignal)/fsTestSignal+0.1); 0107 %Convolve with de-equalized HRIR and play back 0108 supdeq_listen(denseHRTFdataset_sh,testSignal,[azPlayback,elPlayback]); 0109 0110 %% (9) - Optional: Technical Evaluation 0111 0112 %Perform test of sagittal plane localization with default spatial sampling 0113 %grid 0114 spLocalization = supdeq_testSPlocalization(denseHRTFdataset_sh, referenceHRTFdataset); 0115 0116 %Perform test of horizontal plane localizatoin with default spatial 0117 %sampling grid 0118 hpLocalization = supdeq_testHPlocalization(denseHRTFdataset_sh, referenceHRTFdataset); 0119 0120 %Calculate spectral differences with default spatial sampling grid and 0121 %make simple plot 0122 specDifferences = supdeq_calcSpectralDifference(denseHRTFdataset_sh, referenceHRTFdataset,[],[],false,true); 0123 0124 %Calculate spectral differences in auditory bands with default spatial 0125 %sampling grid and make simple plot 0126 specDifferences_auditoryBands = supdeq_calcSpectralDifference(denseHRTFdataset_sh, referenceHRTFdataset,[],[],true,true); 0127 0128 %Calculate energy per order N and make simple plot 0129 energyPerOrder = supdeq_calcEnergyPerOrder(denseHRTFdataset_sh,true);