Onur, you could simply take the derivative of the signal, provided that it is somewhat well behaved. See example below:
t = 0:0.1:20;
x = square(t,5) + 0.1*rand(1,length(t)); % noisy square wave
val = +1; % deriv treshold value
num = length(find(diff(x)>val)); % data points with deriv > threshold
disp(num)
plot(t,x)
