Synthesizing Radar and Communications Waveforms
As previously noted, the occupied frequency spectrum is crowded, and transmitting sources such as communications
systems, radio, and navigation systems all compete for spectrum. To create a test scenario, consider the following
diverse set of waveform types:
• Rectangular
• Linear frequency modulation (LFM)
• Barker code
• Gaussian frequency shift keying (GFSK)
• Continuous phase frequency shift keying (CPFSK)
• Broadcast frequency modulation (B-FM)
• Double sideband amplitude modulation (DSB-AM)
• Single sideband amplitude modulation (SSB-AM)
With these waveforms defined, you can programmatically generate large numbers of I/Q signals for each modulation
type. Each waveform has unique parameters, and the resulting signals are perturbed with various impairments to
increase the fidelity of the model. For each waveform, the pulse width and repetition frequency are randomly gener
ated. For LFM waveforms, the sweep bandwidth and chirp direction are randomly generated. For Barker waveforms,
the chip width and number are generated randomly. All signals are impaired with white Gaussian noise. In addition,
a frequency offset with a random carrier frequency is applied to each signal. Finally, each signal is passed through a
channel model. In this example, a multipath Rician fading channel is implemented, but other models are available
and could be used instead.
The data is labeled as it generated in preparation to feed the training network.
Feature Extraction Using Time-Frequency Techniques
To improve the classification performance of learning algorithms, a common approach is to input extracted features
in place of the original signal data. The features provide a representation of the input data that makes it easier for a
classification algorithm to discriminate across the classes.
In practical applications, many signals are nonstationary. This means that their frequency-domain representation
changes over time. One useful technique to extract features is the Short-Time Fourier Transform (STFT). The STFT
breaks down the signal into overlapping segments, and then applies a Fourier Transform to each segment. This
results in a time-frequency representation of the signal, which can be used as input to a classification algorithm.
Another technique is the Continuous Wavelet Transform (CWT). The CWT is similar to the STFT, but it uses a
wavelet function to analyze the signal. This results in a time-frequency representation of the signal that is more
sensitive to changes in the signal over time.
In addition to the STFT and CWT, other techniques such as the Hilbert Transform and the Wigner-Ville Distribu-
tion can also be used to extract features from nonstationary signals.
The choice of feature extraction technique depends on the specific application and the characteristics of the signal.
In general, the goal is to extract features that are informative and discriminative, and that can be used to improve the
classification performance of the learning algorithm.
Target Classification Using Radar Returns
Target classification is the process of identifying the characteristics of a target based on the radar returns it
produces. This can be a challenging task, especially in environments with clutter and interference.
One approach to target classification is to use a deep learning network to classify the radar returns. The network can
be trained on a dataset of labeled radar returns, where each return is associated with a specific target characteristic.
The network can then be used to classify new, unseen radar returns.
To generate the training data for the network, you can use a combination of simulation and modeling. The simulation
can be used to model the radar returns from a variety of targets, and the modeling can be used to add realistic
impairments to the returns.
The network can be trained on the generated data, and then used to classify new radar returns. This approach can
be used to classify targets based on a variety of characteristics, such as size, shape, and material.
In addition to deep learning networks, other techniques such as machine learning and statistical analysis can also be
used for target classification.
The choice of technique depends on the specific application and the characteristics of the target. In general, the goal
is to develop a system that can accurately classify targets based on their radar returns.
Example 1: Modulation Identification
In this example, we use a deep learning network to identify the modulation type of a received signal. The network
is trained on a dataset of labeled signals, where each signal is associated with a specific modulation type.
The network is then used to classify new, unseen signals. The results show that the network is able to accurately
identify the modulation type of the signals.
Example 2: Target Classification
In this example, we use a deep learning network to classify the radar returns from a variety of targets. The network is
trained on a dataset of labeled returns, where each return is associated with a specific target characteristic.
The network is then used to classify new, unseen returns. The results show that the network is able to accurately
classify the targets based on their radar returns.
Example 3: Object Classification
In this example, we use a deep learning network to classify the objects in a scene based on their radar returns. The
network is trained on a dataset of labeled returns, where each return is associated with a specific object characteristic.
The network is then used to classify new, unseen returns. The results show that the network is able to accurately
classify the objects based on their radar returns.
Conclusion
In this white paper, we have demonstrated how to use MATLAB to synthesize and label radar and communications
waveforms, and to train deep learning networks for modulation identification and target classification.
We have shown that the use of synthetic data can be a powerful tool for training deep learning networks, and that
the resulting networks can be used to classify a wide range of signals and targets.
We have also demonstrated the use of time-frequency techniques for feature extraction, and the use of deep learning
networks for target classification.
The examples provided in this white paper show the potential of these techniques for a variety of applications, and
demonstrate the power of MATLAB for synthesizing and analyzing radar and communications waveforms.
