Introduction to Pattern Recognition: A Matlab Approach

6.3 Recognition and Training 153

Before we proceed, each string will be converted to a sequence of symbol IDs and stored in a ne w cell

array, NumericData. This is convenient because, from an implementation point of view, it is easier to

work with numbers. Assuming that 1 stands for head s and 2 for tails, type

L=length(TrainingData);

for i=1:L

for j=1:length(TrainingData{i})

if TrainingData{i}(j)=='H'

NumericData{i}(j)=1;

else

NumericData{i}(j)=2;

end

end

end

Next, initialize the HMM with the first set of values. Type

pi_init_1 = [0.6 0.4]';

A_init_1 = [0.6 0.4; 0.6 0.4];

B_init_1 = [0.6 0.2; 0.4 0.8];

Then use function MultSeqTrainDoHMMBWsc to train the HMM ...