CSP Estimators: The FFT Accumulation Method

Let’s look at another spectral correlation function estimator: the FFT Accumulation Method (FAM). This estimator is in the time-smoothing category, is exhaustive in that it is designed to compute estimates of the spectral correlation function over its entire principal domain, and is efficient, so that it is a competitor to the Strip Spectral Correlation Analyzer (SSCA) method. I implemented my version of the FAM by using the paper by Roberts et al (The Literature [R4]). If you follow the equations closely, you can successfully implement the estimator from that paper. The tricky part, as with the SSCA, is correctly associating the outputs of the coded equations to their proper $\displaystyle (f, \alpha)$ values.

Can a Machine Learn the Fourier Transform?

(Update: See Part 2 of this post at this link.)

Or any transform for that matter. Or the power spectrum? Autocorrelation function? Cyclic moment? Cyclic cumulant?

I ask because the Machine Learners want to do away with what they call Expert Features in multiple areas involving classification, such as modulation recognition, image classification, facial recognition, etc. The idea is to train the machine (and by machine they seem to almost always mean an artificial neural network, or just neural network for short) by applying labeled data (supervised learning) where the data is the raw data involved in the classification application area. For us, here at the CSP Blog, that means complex-valued data samples obtained through standard RF signal reception techniques. In other words, the samples that we start with in all of our CSP algorithms, such as the frequency-smoothing method, the time-smoothing method, the strip spectral correlation analyzer, the cycle detectors, the time-delay estimators, automatic spectral segmentation, etc.

This is an interesting and potentially valuable line of inquiry, even if it does lead to the superfluousness of my work and the CSP Blog itself. Oh well, gotta face reality.

So can we start with complex samples (commonly called “I-Q samples”, which is short for “inphase and quadrature samples”) corresponding to labeled examples of the involved classes (BPSK, QPSK, AM, FM, etc.) and end up with a classifier with performance that exceeds that of the best Expert Feature classifier? From my point of view, that means that the machine has to learn cyclic cumulants or something even better. I have a hard time imagining something better (that is just a statement about my mental limitations, not about what might exist in the world), so I shift to asking Can a Machine Learn the Cyclic Cumulant?

CSP Blog Highlights

Welcome to the CSP Blog!

To help new readers, I’m supplying here links to the posts that have gotten the most attention over the lifetime of the Blog. Omitted from this list are the more esoteric topics as well as the posts that comment on the engineering literature.

Second-Order Estimator Verification Guide

In this post I provide some tools for the do-it-yourself CSP practitioner. One of the goals of this blog is to help new CSP researchers and students to write their own estimators and algorithms. This post contains some spectral correlation function estimates and numerically evaluated formulas that can be compared to those produced by anybody’s code.

The signal of interest is, of course, our rectangular-pulse BPSK signal with symbol rate $0.1$ (normalized frequency units) and carrier offset $0.05$. You can download a MATLAB script for creating such a signal here.

The formula for the SCF for a textbook BPSK signal is published in several places (The Literature [R47], My Papers [6]) and depends mainly on the Fourier transform of the pulse function used by the textbook signal.

We’ll compare the numerically evaluated formula with estimates produced by my version of the frequency-smoothing method (FSM). The FSM estimates and the theoretical functions are contained in a MATLAB mat file here. (I had to change the extension of the mat file from .mat to .doc to allow posting it to WordPress.) In all the results shown here and that you can download, the processed data-block length is $65536$ samples and the FSM smoothing width is $0.02$ Hz. A rectangular smoothing window is used. For all cycle frequencies except zero (non-conjugate), a zero-padding factor of two is used in the FSM.