track_audio_pf.m

clear;
close all;
clc;

% To be modified


% % % % % %% Request interaural time differences (ITDs)
% % % % % requests = {'itd'};
% % % % %
% % % % % % Parameters of the auditory filterbank processor
% % % % % fb_type       = 'gammatone';
% % % % % fb_lowFreqHz  = 80;
% % % % % fb_highFreqHz = 8000;
% % % % % fb_nChannels  = 32;
% % % % %
% % % % % % Parameters of innerhaircell processor
% % % % % ihc_method    = 'dau';
% % % % %
% % % % % % Parameters of crosscorrelation processor
% % % % % cc_wSizeSec  = 0.02;
% % % % % cc_hSizeSec  = 0.01;
% % % % % cc_wname     = 'hann';
% % % % %
% % % % % % Summary of parameters
% % % % % par = genParStruct('fb_type',fb_type,'fb_lowFreqHz',fb_lowFreqHz,...
% % % % %     'fb_highFreqHz',fb_highFreqHz,'fb_nChannels',fb_nChannels,...
% % % % %     'ihc_method',ihc_method,'cc_wSizeSec',cc_wSizeSec,...
% % % % %     'cc_hSizeSec',cc_hSizeSec,'cc_wname',cc_wname);

%% Setup objects
% Initialize localization models using braodband and subband settings
dObj   = dataObject([],fsHz,10,2);

% Settings for subband approach
par_sub = genParStruct('cc_bBroadband',0,'cc_wSizeSec',winSec,...
    'cc_hSizeSec',winSec/2,'cc_maxDelaySec',1.25e-3,...
    'fb_lowFreqHz',fLowHz,'fb_highFreqHz',fHighHz,...
    'fb_nERBs',1,'ihc_method','none',...
    'loc_NSources',nSpeakers(hh));

% Initialize localization models using braodband and subband settings
mObj  = manager(dObj,'localization',par_sub);

%% Particle filter parameters
sigma_w = 1; % State noise std
sigma_v = 1; % Measurement noise std
fs = 24414;
chunks = 2048;
dt = chunks/fs;

Q = [1/4*dt^4, 1/2*dt^3; 1/2*dt^3, dt^2]*sigma_w;                       % Process noise covariance
R = 1;                                  % Measurement noise covariance
x = [0; 0];                             % Initial state

A = [1, dt; 0, 1];
Npart = 100;
f = @(x) A*x + sigma_w*diag(Q)*randn(1,size(x,2));
g_density = @(y,x) -0.5/sigma_v * (y-x(1,:)).^2; %% This assumes gaussian noise, try to change to Laplacian noise for robustness

%% Initialization


figure(1)

N = 100;

% Initialize posterior mean and covariance
posteriorMean = zeros(1, N);
posteriorCovariance = zeros(size(A, 1), size(A, 1), N);


% =======================================================
% Main loop - Perform localization and tracking
% =======================================================
for l = 1:N
    audio = get_audio();
    
    % Request processing
    mObj.processSignal(audio);
    azimEst = dObj.localization{1}.Data(end,1); % There might be an issue with several sources here!
    % Perform Kalman filter prediction and update, TODO: consider changing this
    % crappy filter for a PF
    [xp, w, expw] = pf( azimEst, xp, w, expw, g_density, f);
%     [~,I] = max(expw);
    xh = sum(xp(1,:).*expw);
    posteriorMean(l) = xh;
end

% Plot measurements
subplot(2, nFiles / 2, k);
timeAxis = linspace(0, nSamples / fsHz, nFrames);
plot(timeAxis, measuredLocations, 'x', 'LineWidth', 2);
axis([0, nSamples / fsHz, -90, 90]);
xlabel('Time / s');
ylabel('Azimuth / deg');
grid on; hold on;
plot(timeAxis, posteriorMean(1, :), 'g', 'LineWidth', 2);

% Plot ground truth
plot(timeAxis, gtTrajectory, 'r--', 'LineWidth', 2);
legend('Measurements', 'Estimated trajectory', 'Ground truth');

% Compute RMSE
rmse = sqrt(sum((posteriorMean(1, :) - gtTrajectory).^2) ./ nFrames);

if ~strcmpi(noiseType, 'none')
    title([upper(soundType), ', ', upper(noiseType), ' NOISE AT ', ...
        num2str(snr), ' dB SNR, ', 'RMSE: ', num2str(rmse), '°']);
else
    title([upper(soundType), ', NO NOISE, ', 'RMSE: ', ...
        num2str(rmse), '°']);
end