% SUpDEq - Spatial Upsampling by Directional Equalization
function eqDataset_limited = supdeq_limitEqDataset(eqDataset, Nmax, radius, safetyMargin, fade)
This function can be used to limit the equalization in frequency domain
dependent on Nmax of the sparse HRTF set. Below the specific aliasing
frequency fA, the eqDataset will be set to 0, and thus the subsequent
equalization will have no effect in this frequency range, resulting in
normal SH interpolation below fA. N = 0 is not affected by the limiting!
IMPORTANT (1): If limitEq is applied for equalization using supdeq_eq,
it must also be applied for de-equalization using supdeq_deq!
IMPORTANT (2): As fA depends on the radius, and the transition between
normal SH interpolation and SUpDEq processing should be the same in the
equalization and de-equalization, the radius (of the eqDataset) can be
passed here. Thus, if a different radius is applied for de-equalization,
the limiting of the deqDataset should be done based on the radius of the
eqDataset.
Output:
eqDataset_limited - Limited eqDataset
Input:
eqDataset - eqDataset struct obtained with supdeq_getEqDataset
Nmax - Maximum spatial order N of the sparse HRTF set
radius - Optionally, the radius can be passed if fA should
be calculated based on another radius than
specified in the eqDataset struct. This can be
usefull, if the deqDataset has a different radius,
but the limiting should affect the dataset in the
exact same frequency range (see important hints).
If the radius is not passed, the values specified
in the eqDataset struct will be applied.
safteyMargin - Boolean to choose if safety margin (third-octave
below fA) is applied or not.
Default: true
fade - Boolean to choose whether to apply a fade from fA
to fA/(2^1/3) or not
Default: false (to be compatabile with Arend et al. 2021)
Only applies if safteyMargin = true
Dependencies: -
(C) 2019 by JMA, Johannes M. Arend
CP, Christoph Pörschmann
TH Köln - University of Applied Sciences
Institute of Communications Engineering
Department of Acoustics and Audio Signal Processing0001 %% SUpDEq - Spatial Upsampling by Directional Equalization 0002 % 0003 % function eqDataset_limited = supdeq_limitEqDataset(eqDataset, Nmax, radius, safetyMargin, fade) 0004 % 0005 % This function can be used to limit the equalization in frequency domain 0006 % dependent on Nmax of the sparse HRTF set. Below the specific aliasing 0007 % frequency fA, the eqDataset will be set to 0, and thus the subsequent 0008 % equalization will have no effect in this frequency range, resulting in 0009 % normal SH interpolation below fA. N = 0 is not affected by the limiting! 0010 % 0011 % IMPORTANT (1): If limitEq is applied for equalization using supdeq_eq, 0012 % it must also be applied for de-equalization using supdeq_deq! 0013 % 0014 % IMPORTANT (2): As fA depends on the radius, and the transition between 0015 % normal SH interpolation and SUpDEq processing should be the same in the 0016 % equalization and de-equalization, the radius (of the eqDataset) can be 0017 % passed here. Thus, if a different radius is applied for de-equalization, 0018 % the limiting of the deqDataset should be done based on the radius of the 0019 % eqDataset. 0020 % 0021 % Output: 0022 % eqDataset_limited - Limited eqDataset 0023 % 0024 % Input: 0025 % eqDataset - eqDataset struct obtained with supdeq_getEqDataset 0026 % Nmax - Maximum spatial order N of the sparse HRTF set 0027 % radius - Optionally, the radius can be passed if fA should 0028 % be calculated based on another radius than 0029 % specified in the eqDataset struct. This can be 0030 % usefull, if the deqDataset has a different radius, 0031 % but the limiting should affect the dataset in the 0032 % exact same frequency range (see important hints). 0033 % If the radius is not passed, the values specified 0034 % in the eqDataset struct will be applied. 0035 % safteyMargin - Boolean to choose if safety margin (third-octave 0036 % below fA) is applied or not. 0037 % Default: true 0038 % fade - Boolean to choose whether to apply a fade from fA 0039 % to fA/(2^1/3) or not 0040 % Default: false (to be compatabile with Arend et al. 2021) 0041 % Only applies if safteyMargin = true 0042 % 0043 % Dependencies: - 0044 % 0045 % (C) 2019 by JMA, Johannes M. Arend 0046 % CP, Christoph Pörschmann 0047 % TH Köln - University of Applied Sciences 0048 % Institute of Communications Engineering 0049 % Department of Acoustics and Audio Signal Processing 0050 0051 function eqDataset_limited = supdeq_limitEqDataset(eqDataset, Nmax, radius, safetyMargin, fade) 0052 0053 if nargin < 3 || isempty(radius) 0054 radius = eqDataset.radius; 0055 end 0056 0057 if nargin < 4 || isempty(safetyMargin) 0058 safetyMargin = true; 0059 end 0060 0061 if nargin < 5 || isempty(fade) 0062 fade = false; 0063 end 0064 0065 if isfield(eqDataset,'limited') 0066 if eqDataset.limited 0067 error('Input eqDataset already limited!'); 0068 end 0069 end 0070 0071 %Just a variable to check if equalizing should be applied 0072 noEq = false; 0073 0074 %Get aliasing frequency fA (N ~ kr) 0075 fA = Nmax * eqDataset.c/ (2*pi*radius); 0076 %Check if fA is >= fs/2 0077 if fA >= eqDataset.f(end) 0078 noEq = true; 0079 end 0080 0081 if safetyMargin 0082 %Lower by 1/3 octave to ensure distance to aliasing frequency 0083 fA_shift = fA/2^(1/3); 0084 else 0085 fA_shift = fA; 0086 end 0087 %Get corresponding index of frequency bins 0088 [~,fA_shift_bin] = min(abs(eqDataset.f-fA_shift)); 0089 0090 %Get eqDataset_limited 0091 eqDataset_limited = eqDataset; 0092 0093 if ~noEq %If SUpDEq should be partly applied 0094 0095 if safetyMargin && ~fade %Without fade 0096 %Set all SH coefficients for N > 0 to 0 0097 eqDataset_limited.Hl_nm(2:end,1:fA_shift_bin)=0; 0098 eqDataset_limited.Hr_nm(2:end,1:fA_shift_bin)=0; 0099 end 0100 0101 if safetyMargin && fade %With fade from fA_shift to fA 0102 [~,fA_bin] = min(abs(eqDataset.f-fA)); 0103 0104 ramp = linspace(0,1,abs(fA_bin-fA_shift_bin)+1); 0105 0106 eqDataset_limited.Hl_nm(2:end,1:fA_shift_bin-1)=0; 0107 eqDataset_limited.Hl_nm(2:end,fA_shift_bin:fA_bin)=eqDataset_limited.Hl_nm(2:end,fA_shift_bin:fA_bin).*ramp; 0108 eqDataset_limited.Hr_nm(2:end,1:fA_shift_bin-1)=0; 0109 eqDataset_limited.Hr_nm(2:end,fA_shift_bin:fA_bin)=eqDataset_limited.Hr_nm(2:end,fA_shift_bin:fA_bin).*ramp; 0110 end 0111 0112 else %If fA is > fs/2 and SUpDEq does not need to be applied 0113 %Note: Actually, in this case SH interpolation without any equalization is 0114 %perfectly fine anyway ;-). 0115 0116 %Set all SH coefficients for N > 0 to 0 0117 eqDataset_limited.Hl_nm(2:end,1:end)=0; 0118 eqDataset_limited.Hr_nm(2:end,1:end)=0; 0119 end 0120 0121 %Write some infos in new struct eqDataset_limited 0122 eqDataset_limited.limited = true; 0123 eqDataset_limited.limitInfo.Nmax = Nmax; 0124 eqDataset_limited.limitInfo.appliedRadius = radius; 0125 eqDataset_limited.limitInfo.fA = fA; 0126 if ~noEq 0127 eqDataset_limited.limitInfo.fA_shift = fA_shift; 0128 if safetyMargin && fade 0129 eqDataset_limited.limitInfo.fade = true; 0130 end 0131 else 0132 eqDataset_limited.limitInfo.noEq = noEq; 0133 end 0134 0135 fprintf('eqDataset limited depending on Nmax = %d\n',Nmax);