| 1 | function [adaptFilterOut, convergeIdx] = adaptiveCMA(rSampled) |
| 2 | %% adaptive filter |
| 3 | %% CMA |
| 4 | taps = 19; % ODD taps |
| 5 | hxx = zeros(taps, 1); |
| 6 | %% hxx: real indices -K, ..., 0, ..., K. K = floor(taps/2) |
| 7 | %% MATLAB indices 1 1+K taps |
| 8 | %% initialize hxx, hxx[0] = 1, hxx[k] = hxx[-k] = 0 |
| 9 | hxx(ceil(taps/2)) = 1; |
| 10 | |
| 11 | mu = 1e-3; |
| 12 | numSymbs = length(rSampled); |
| 13 | |
| 14 | %% Check average energy of symbols |
| 15 | rSampledUnitEnergy = normalizeEnergy(rSampled, numSymbs, 1); |
| 16 | |
| 17 | adaptFilterOut = zeros(numSymbs, 1); |
| 18 | converged = 0; |
| 19 | convergeCount = 0; |
| 20 | convergeIdx = Inf; |
| 21 | |
| 22 | for it = 1:numSymbs |
| 23 | if it <= (taps - 1) / 2; |
| 24 | xp = [zeros((taps - 1) / 2 - it + 1, 1); rSampledUnitEnergy(1:it + (taps - 1) / 2)]; |
| 25 | elseif it + (taps - 1) / 2 > numSymbs |
| 26 | xp = [rSampledUnitEnergy(it - (taps - 1) / 2 : end); zeros(it + (taps - 1) / 2 - numSymbs, 1)]; |
| 27 | else |
| 28 | xp = rSampledUnitEnergy(it - (taps - 1) / 2 : it + (taps - 1) / 2); |
| 29 | end |
| 30 | |
| 31 | xout = sum(hxx .* xp); |
| 32 | ex = 1 - abs(xout) ^ 2; |
| 33 | |
| 34 | if abs(ex) < 0.01 |
| 35 | convergeCount = convergeCount + 1; |
| 36 | else |
| 37 | convergeCount = 0; |
| 38 | end |
| 39 | if ~converged && convergeCount >= 10 |
| 40 | converged = 1; |
| 41 | convergeIdx = it; |
| 42 | end |
| 43 | |
| 44 | adaptFilterOut(it) = xout; |
| 45 | |
| 46 | hxx = hxx + mu * ex * xout * conj(xp); |
| 47 | end |
| 48 | |
| 49 | %{ |
| 50 | %% try MATLAB builtin equalizer |
| 51 | alg = cma(mu); |
| 52 | eqObj = lineareq(taps, alg); |
| 53 | eqObj.Weights((taps + 1) / 2) = 1; |
| 54 | rPadded = [rSampledUnitEnergy; zeros((taps - 1) / 2, 1)]; |
| 55 | matlabEq = equalize(eqObj, rPadded); |
| 56 | matlabEq = matlabEq((taps + 1) / 2 : end); |
| 57 | %} |
| 58 | end |