I am trying to find a 'threshold value' for separation of the noise components and the signal components in the received signal r1. How can I do that?

I want to set all frequency components below that threshold to 0 and store this as a modified filtered spectrum, ( take care for the symmetry of the complex spectrum for a real signal) and inverse FFT this modified spectrum to obtain the corresponding real signal rx. The sequence of tones rx is a received signal after noise suppression, so DTMF codes can be detected. How can I do that?? Please help me

Signal1= 0.5*[ sin(2*pi*f1(5)*t)+sin(2*pi*f2(5)*t)]; 

r1= Signal1+ randn(size(t)); % Random noise signal 

R1= fft(r1, 512); % fft of S1
p= R5.* conj(R5)/512; % power spectrum
f= 1000*(0:256)/512; % frequency scale

[Image Analyst]

I fixed it for you, but read this. And give us code that runs. What is f1, f2, t, and R5?