Adding MASA SRIR examples. Includes SRIR analysis, SRIR normalization and SRIR...

% Example script on analysis and normalization of SRIR (Spatial Room

% Impulse Responses) from recordings.

% This script also shows how IVAS MASA format SRIR can be created and used for listening test processing.

%

% In order to fully run the script you need also

% IVAS_rend         IVAS renderer 

% masaAnalyzer      IVAS MASA analyzer tool

%

% Windows binaries are included, but for Linux/Mac you need to add your own

%

%

% Author: Anssi Rämö (Nokia Techonologies)

% Date: 10.4.2023

%

%

%

clear

fs=48000;         % Sampling frequency

ms = fs/1000;     % millisecond in samples

DC=1;             % Preprocessing (DC-Block filtering may be useful, if lot of low frequency brum and recorded without high-pass filter)

PeakPos=2;        % in ms (SRIR peak position, for the closest channel)

IR_Len=200;       % in ms

FadeInOut=1;      % Post-processing of SRIR

FadeInLength=24;  % in samples. Sinusoidal FadeIn -length

FadeOutLength=60; % in samples. Cosine FadeOut -length

% Test signal that has been used for recordings

target=audioread('ExampleInput\TargetMLS.wav');target=target(:);

% The recording(s) made in a room using the test signal

AmbeoFoA=audioread('ExampleInput\SennheiserAmbeo_FoA_Pos3.wav');

ExtM5=audioread('ExampleInput\RodeExtM5_Cardioid_Pos3.wav');

% IVAS MASA format may utilise SRIR responses of form FoA + Ext transport (mono/stereo)

% FoA signal is created by VST-plugin while recording. If processing delay is not

% compensated during recording it has to be conmpensated in analysis

% We have found out following delays for different ambisonic microphones

% These depend on the DAW used and their settings.

RodeNTSF1_DelayComp=1312+1+1024;      % Delay compensation for Rode NT-SF1 VST plugin FoA analysis

SennheiserAmbeo_DelayComp=1312+1+128; % Delay compensation for Sennheiser Ambeo VST plugin FoA analysis

VoyageAudio_DelayComp=1312+1+512; %   % Delay compensation for Voyage Audio plugin HoA2 analysis

% Aling FoA + Ext audio signals for SRIR calculation

measurement=[AmbeoFoA(SennheiserAmbeo_DelayComp+1:end,:) ExtM5(1:end-SennheiserAmbeo_DelayComp,:)];

% Pre-process (DC-block, highpass filtering 30Hz) recorded signal, if needed

if (DC)

    measurement=DCBlock(measurement);

end

% Do the SRIR analysis

chn=min(size(measurement)); % Number of channels to analyze

len = length(target);

for i = 1:chn

    Y=fft(measurement(:,i),(length(measurement)+size(target,2)-1));

    H=fft(target,(length(measurement)+size(target,2)-1));

    G=Y./(H);

    SRIR_long(:,i)=ifft(G,'symmetric');

    [null,latencies(i)] = max(SRIR_long(1:fs,i)); % Peak has to be within first second, if not adjust your measurement signal range

end

latency = min(latencies); % Put the closest microphone signal to PeakPos

SRIR_short = SRIR_long(latency-PeakPos*ms+1:latency+(IR_Len-PeakPos)*ms,:);

% FadeIn, FadeOut

if (FadeInOut==1)

    inwin=sin(pi*(0:FadeInLength-1)/(2*FadeInLength-2)).^2;

    FadeIn=repmat(inwin',[1 chn]);

    SRIR_short(1:FadeInLength,:)=SRIR_short(1:FadeInLength,:).*FadeIn;

    outwin=cos(pi*(0:FadeOutLength-1)/(2*FadeOutLength-2)).^2;

    FadeOut=repmat(outwin',[1 chn]);

    SRIR_short(end-FadeOutLength+1:end,:)=SRIR_short(end-FadeOutLength+1:end,:).*FadeOut;

end

%plot(SRIR_short)

% Do SRIR normalization

% IVAS MASA input format FoA + Ext parts are best to be normalized separately

% for the FoA/HoA part and transport signals.

% Test signal with short sequences of male & female speech and white & pink noise. All normalized to -26 dbOV

test26=audioread('ExampleInput\Test-26dbOV.wav');

Tmp_Talker=zeros([length(test26) chn]);

% Make initial SRIR scaling something reasonable

SRIR_short=0.2*SRIR_short/max(max(abs(SRIR_short))); 

% Filter test signal with SRIR vectors

for t=1:chn

    Tmp_Talker(:,t)=filter(SRIR_short(:,t),1,test26);

end

% Normalize FoA part based on the binaural rendered signal.

% Render FoA to binaural with IVAS_rend-tool

audiowrite('FoA.wav',Tmp_Talker(:,1:4),48000,'BitsPerSample',16);

eval('!IVAS_rend.exe -i FoA.wav -if FOA -o BIN.wav -of BINAURAL');

bin=audioread('BIN.wav');

% Calculate BS.1770 loudness and scale to target level (-26 dBOv)

[level gain]=bs1770(bin,-26) 

SRIR_short(:,1:4)=gain*SRIR_short(:,1:4);

% Normalize transports, directly from the Ext-transport stereo signal

[leveltr gaintr]=bs1770(Tmp_Talker(:,5:6),-26)

SRIR_short(:,5:6)=gaintr*SRIR_short(:,5:6);

% Write normalized SRIR to a file

audiowrite('Nokia_LargeRoom_Pos3_FoA+Ext_6hcn.wav',SRIR_short,48000,'BitsPerSample',16);

% Render spatialized mono signal with stereo MASA FoA+Ext processing

if (1)

    SRIR=audioread('Nokia_LargeRoom_Pos3_FoA+Ext_6hcn.wav');

    speech=audioread('ExampleInput\FinnishMale_mono.wav');

    SRIR_speech=zeros([length(speech) 6]);

    for t=1:6

        SRIR_speech(:,t)=filter(SRIR(:,t),1,speech(:,1)); % Anssi

    end

    audiowrite('FoAExt.wav',SRIR_speech,48000,'BitsPerSample',16);

    WavToPCM('FoAExt.wav');

    % Analyze FoA + stereo Ext signal with masaAnalyzer and create stereo.pcm and MASA metadata.

    !masaAnalyzer -stereo -1dir -foaext FoAExt.pcm stereo.pcm 1dir.met

    eval(['!IVAS_rend.exe -fs 48 -if MASA2 -im 1Dir.met -i stereo.pcm -of BINAURAL -o FinMale_Pos3_BIN.wav']);

end

 No newline at end of file

function y = DCBlock ( x )

% DC-block filtering. Default is 30Hz (-3 dB)

% a = 1-2^-11 = 0.99951171875  -3dB cut-off 3.8Hz, 20Hz atten. -0.004dB

% a = 1-2^-9  = 0.998046875    -3dB cut-off 15Hz, 20Hz atten. -1.9 dB

% a = 1-2^-8  = 0.9961         -3dB cut-off 30Hz, 20Hz atten. -5.1 dB

% a = 1-2^-7  = 0.9922         -3dB cut-off 60Hz, 20Hz atten. -10 dB

% a = 1-2^-6  ~= 0.985         -3dB cut-off 120Hz, 20Hz atten. -15? dB (ITU-T STL p50mnru default)

c = 1-2^-8;         % c - coefficient

b = [1 -1];	 	 % filter coefficient b

a = [1 -c];	 	 % filter coefficient a

y = filter (b,a,x);

end

 No newline at end of file