detect.m

MIMO and VBLAST( please see the documentation)

function varargout = detect(sig,lt_data,lt_state,ch_coefs,samples,...
  smplper,gain,pulse,varargin)

%DETECT Multidimensional Viterbi hard decision detector.
%   DATA_EST = DETECT(SIG,LT_DATA,LT_STATE,CH_COEFS,SAMPLES,SMPLPER
%   ,GAIN,PULSE) returns the data estimations based on look-up
%   tables LT_DATA and LT_STATE. An external function BRMET is called
%   during the computation to evaluate branch metric. CH_COEFS includes
%   a channel complex path fadings; SAMPLES corresponds to the samples 
%   per symbol; GAIN corresponds to the constellation expansion factor
%   computed by the SCALE function and PULSE contains the samples of
%   modulation impulse sampled at frequency SMPLFREQ = 1 / SMPLPER.
%
%   [DATA_EST,STATE_EST] = DETECT(...) also returns matrix with trellis
%   states of most probable path. This matrix is required when the
%   decoding process is displayed with DISPTRELL function.
%
%   DATA_EST = DETECT(...,'EchoOn') toggles function echo on. When
%   'EchoOn' option is chosen than a summary in form below is displayed.
%
%      DETECT:    Performing data detection. This may take a while.
%                 Please wait...
%                 Total decoding complexity -> ? steps.
%                 Total elapsed time -> ? hrs, ? min, ? sec.
%
%   See also BRMET, DISPTRELLIS, MAKEPULSE, SCALE.

%   Copyright 2001-2002 Kamil Anis, [e-mail]
%   Dept. of Radioelectronics, 
%   Faculty of Electrical Engineering
%   Czech Technical University in Pragu
%   $Revision: 2.0 $  $Date: 2002/10/23 17:33:28 $
%   --
%   <additional stuff should go here>

if (isempty(varargin) == 0) & (varargin{end} == 'EchoOn')
	[indent,gap,name] =  iprompt('DETECT:');
	
	disp(' ');
	disp([name,gap,'Performing data detection. This may take a while.']);
	disp([indent,'Please wait...']);
end

[sig_length,space_dim,frames] = size(sig);
[s,md,space_dim] = size(lt_data);
load qam16.txt;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% BODY BEGIN %%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

tic;
for k = 1:frames
	% passing the signal through matched filter
  clear sig_mf;
	for j = 1:space_dim
    sig_mf(:,j) = smplper * conv(pulse,sig(:,j,k)) / gain;
	end

  % omiting the sig begining to prevent of insecure decisions
	delay = length(pulse);
	sig_mf = sig_mf(delay:end-delay,:,:);

	% sampling at symbol period (Ns in Dt)
	sig_down = downsample(sig_mf,samples);
	[step_final,foo] = size(sig_down);

	% running multi-dimensional Viterbi algorithm
	% make all starting paths unprobable except for the correct one
	metric(1,2:s) = realmax;

	for l = 1:step_final
		for j = 1:s % current j
		% finding all previous states s_pre leads to current sate j
		[s_pre,foo] = find(lt_state == j);

		% determining a pair position relevant to the state j
		% {1,2,3,4,5,6,7,8} -> {1,2,3,4,1,2,3,4}
		pos = mod(j - 1,md) + 1;

		% picking-up the pairs corresponding to each of s_pre states
		data_test = lt_data(s_pre,pos,:);
		data_test = reshape(data_test,[md space_dim]);

		% mapping pairs to appropriate constellation
		if md == 16 % 16QAM
			for r = 1:space_dim
				k1(:,r) = qam16(data_test(:,r) + 1,1);
				k2(:,r) = qam16(data_test(:,r) + 1,2);
			end

			q_test = (2 * k1 - md - 1) - i * (2 * k2 - md - 1);

		else % 4,8PSK
			expr = i * 2 * pi / md;
			q_test = exp(expr * data_test);
		end

		% evaluating branch metric
		metric_d = brmet(sig_down(l,:),q_test,ch_coefs(:,:,k));

		% adds the data_test metrices to the previous states
		metric_md = metric(l,s_pre)' + metric_d;

		% choosing a metric with lowest accumulated value
		[metric_min,metric_pos] = min(metric_md);

		% and storing it's value to the matrix of metrices
		metric(l + 1,j) = metric_min;

		% creates a states matrix of s_pre (with lowest metric)
		vit_state(l + 1,j) = s_pre(metric_pos);

		% creates a matrix of appropriate data_test
		vit_data(l + 1,j) = pos - 1;
		end
	end

	% finding the best path at the trellis end
	[foo,state_best] = min(metric(end,:));
	state_est(step_final + 1) = state_best;

	% back tracking
	for l = step_final:-1:1
		state_est(l) = vit_state(l + 1,state_est(l + 1));
		data_est(l,:,k) = vit_data(l + 1,state_est(l + 1));
	end
end

totaltime = toc;

varargout = {data_est,state_est};

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% BODY END %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

if (isempty(varargin) == 0) & (varargin{end} == 'EchoOn')
	hour = round(totaltime / 3600);
	mins = round(totaltime / 60);
	secs = mod(totaltime,60);
	
	str1 = num2str(frames * sig_length * s);
	str2 = sprintf('%1d',hour);
	str3 = sprintf('%1d',mins);
	str4 = sprintf('%1.1f',secs);
	
	disp([indent,'Total decoding complexity -> ',str1,' steps.']);
	disp([indent,'Total elapsed time -> ',str2,' hrs, ',str3,' min, ',str4,...
        ' sec.']);
	disp(' ');
end