what do you mean by complex sound, data you recorded is microphone data in time domain ?

Yeah real time sound data of drone fly and environmental noises are recorded by four-microphone array.

As its activity to detect drone and most of the data part is drone sound so it consists of harmonics frequencies.

If it has harmonics they could be found out when we analyse the sound, but that does not mean "complex and real time sound" .

as mentioned in the snip you have shared you need to take STFT of the recorded sound and look for peaks in magnitude spectrum, this link could be helpful :-

https://in.mathworks.com/help/signal/ref/spectrogram.html

you need to read about STFT and frequency resolution for this problem.

I designed function of above algorithm but i'm unable to specify the threshold frequency Ft and N.. i just know sampling frequency of drone sound data. Arpit Jain

function f=fre_harmonic(s,fl,fr,ft,N,fs)

%fs: sampling frequency

FFT_s=abs(fft(s,N));

fl1=ceil(fl*fs/N); %Transform Frequency to Points + inf

fr1=floor(fr*fs/N); %Transform Frequency to Points - inf

[v,f(1)]=max(FFT_s(fl1:fr1));

for n=2:N

fl1=ceil((n/(n-1)*f(n-1)-ft)*fs/N);

fr1=floor((n/(n-1)*f(n-1)+ft)*fs/N);

[v,f1]=max(FFT_s(fl1:fr1));

f(n)=f1+fl1;

end

Kindly help me so that i may get able to complete above mentioned process in main question.