It would not be easy to perform your task with vectorized code, so here is code using a for-loop. Let v1 be the indicated row vector of numbers, let NB be an equally long row vector of logicals which are true for non-bold numbers in v1 and false for bold numbers. Let T be the threshold value.
v2 = ((v1>T)&NB)-((v1<T)&NB);
s1 = []; s2 = [];
b = 0;
c = 0;
for ix = 1:length(v1)
if b==1
if v2(ix)==-1, b = -1; s2 = [s2,c]; c = 1;
else, c = c + 1;
end
elseif b==-1
if v2(ix)==1, b = 1; s1 = [s1,c]; c = 1;
else, c = c + 1;
end
else
c = c + 1;
if v2(ix)==1, b = 1;
elseif v2(ix)==-1, b = -1;
end
end
end
if b==1, s2 = [s2,c];
else s1 = [s1,c];
end
There is an ambiguity in your problem as to how to set s1 and s2 if all numbers in v1 are bold. This code puts the count in s1 in that case. Also if a number in v1 is exactly equal to the threshold, then it is treated as though it were a "bold" number.
