Finding threshold values for classification

4 次查看(过去 30 天)
I have been struggling too much on a book question that I know the answer, but cannot find the mathematical solution. The question goes like this:
A device uses a camera which has discovered three spectral peaks that reflect different types of particles. Each spectra follows a normal distribution. The parametric distributions of the three particle classes are N(7,2*2), N(15,5*5), and N(3,5*5). The manufacturers need help to define the thresholds to perform robust classification. They wish to perform a parametric classification of the three classes.
What signal thresholds should they use?
Answer: 4.24 and 10.29
The previous question was similar, but it asked for the minimum distance classification instead of the parametric classification. For this I ploted the peaks as lines and then found the average between those lines and got that the thresholds should be 5 and 11, which is correct:
% N1 = N(7,2*2);
% N2 = N(15,5*5);
% N3 = N(3,5*5);
signal1 = (3+7)/2; % N3-N2
signal2 = (15+7)/2; % N1-N2
figure
hold on
xline(signal2, 'r');
xline(signal1, 'r');
xline(7)
xline(15)
xline(3)
I have been trying to solve it for the parametric classification, but have no idea how to do it. Could someone please point me in the right direction?

回答(1 个)

yanqi liu
yanqi liu 2021-11-20
sir,use for classification,may be use the svm or random forest ……

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by