How to find clustering directions from a dataset?

I have a time-domain dataset that has the form of x (where x(t) = A s(t)). In the transformed (Fourier) domain the dataset is clustered in certain directions which are given by the vectors of matrix A. I am trying to use fuzzy c-means clustering to get these directions (i.e. column vectors forming A) back. How can I do so? (I tried using the built-in function fcm but could not get anything useful for directions). Here's my code

clc
clear;
close all;

fs =500;

t = 0:1/fs:10-1/fs;

s = [sin(2*pi*2*t); sin(2*pi*10*t); sin(2*pi*50*t)];

A = [1  0.5  0.1;
     1 -0.5  0.3;];

x = A*s;

X = (dct(x'))'; 

plot(X(1,:),X(2,:))
hold on
plot(X(1,:),X(2,:),'O')
xlabel('X_1(f)'); ylabel('X_2(f)');

% findcluster

[center,U,obj_fcn] = fcm(X,3);

You can see from the plot that the data tends to cluster in the directions of a1=[1;1], a2=[0.5;-0.5] and a3=[0.1;0.3]