automatic modulation recognition

[dewdropz]

I need to implement an algorithm for automatic modulation recognition using matlab. I am planning to do this by using pattern recognition, I have read about the decision tree approach, it says to "ideniify the constant envelope signals (CW, FM, FSK). PSK sigals are
not considered as constant envelope signals, since in practice
they are band-limited, which leads to a non-constant
envelope. The feature used to identify the envelope variations
is the maximum of the squared Fourier transform of the
normalized signal amplitudes "
How can i implement this technique in matlab, i need help with the code. I also need to calculate the instantaneous freq/phase, etc. I am using a .wav file.

[Taywin]

Err... It's been 2 years since I coded Matlab last time... You are going to create a learning agent by passing some sample data for it to learn using decision tree? If you are going to use decision tree, you must be sure that you can find all factors and each factor is not overlapped. However, I would not suggest you to use decision tree. If you want a simple one, use network algorithm instead. I have never done Matlab with sound file but only image and movie files. :(

[dewdropz]

yes, i believe network would be easier, how did u implement it? is there any chance you can forward me your code, it would be a great help.

[Taywin]

Hmm... I implemented the neural network in Lisp :P Though, the concept of how to do it can be found at wikipedia. You could also look for any AI book which will teach you how to do it. It is a bit higher level than decision tree, but it is at least a simple algorithm to follow.

[pratikawasthi90]

Hi , I have to make simulation project , in that I have to implement Digital modulation recoginition using decition theoratic approach given in this paper " Algorithms for Automatic Modulation Recognition of Communication Signals Asoke K. Nandi, Senior Member, IEEE, and E. E. Azzouz " given in the page 3. I am unble to understand how to get the threshhold values and the the values from the digital signal.

I am in my B.tech final year , please help!!!

[Taywin]

Please give a correct bibliography when you are referencing something. Given only an author name & the article may not be enough to search for the article; besides, the page number you said is wrong. The article was published in "IEEE Transactions on Communications," Vol.46, No.4, April 1998. The page number of the article is from 431-436.

The method to find the threshold is written clearly on page 434 (section IV.A) where it said that you use the minimum value of K(x) where K(x) is a computation from probability of subset A & B. The subsets are the two nonoverlapping of a set of modulation types. Therefore, before you can do anything, you need to divide the set of modulation into 2 nonoverlapping sets. Then you can work from there.