Classifying Signals in Wi-Fi 6 and 5G

Wireless technology has grown quickly, making it very important to use the radio spectrum wisely. Many methods, such as massive MIMO and cognitive radio, are being looked at to help with this issue. One key part of cognitive radio is Spectrum Sensing. This allows devices to understand how the radio spectrum is being used in real-time and make decisions about how to use it better. This article focuses on classifying the signals used in popular wireless communication systems like Wi-Fi 6 and 5G.

#What is OFDM?

Orthogonal Frequency Division Multiplexing (OFDM) is a method used in many modern communication systems. It sends data over multiple smaller sub-channels, which helps to prevent interference and improves reliability. In OFDM systems, data is first converted into symbols, which are then sent over different subcarriers.

Both Wi-Fi 6 and 5G use OFDM. In these systems, bits of information are turned into digital symbols using a technique called quadrature amplitude Modulation (QAM). Many symbols are sent at once, so a single time sample only contains a small part of the information from an OFDM symbol.

#The Challenge with Modulation Classification

Because of the way OFDM works, identifying the modulation type can be very tricky. Traditional methods designed for simpler communication systems don't work well for OFDM signals. A classifier that can accurately identify the modulation of Wi-Fi 6 and 5G signals must do more than just look at raw time-domain samples.

In a regular receiver, the system has access to specific information about the signal being sent. However, a spectrum sensor does not have this luxury. It needs to classify the modulation without knowing the details of the signal, like the sizes of the Fast Fourier Transform (FFT) or the lengths of the Cyclic Prefix (CP). This lack of information makes it hard to identify the modulation scheme.

#How This Study Works

This study presents a solution for classifying the modulation of Wi-Fi 6 and 5G signals. The proposed method does not need any prior knowledge of the signals, such as their carrier frequency or control information. Instead, it relies on the basic structure of OFDM and estimates the necessary parameters like CP length and subcarrier spacing.

The goal is to identify modulation schemes in Wi-Fi 6 signals and the downlink shared channels of 5G. This method works for single-input single-output (SISO) configurations and focuses on frequencies below 7.125 GHz.

To estimate the parameters, the study uses a method called cyclic autocorrelation function (CAF). This technique helps to detect repeated sequences in wireless signals, which allows for accurate parameter estimation.

#Estimating Cyclic Prefix and Subcarrier Spacing

One of the main tasks is to estimate the cyclic prefix length and subcarrier spacing. The CAF technique is used for this purpose. The cyclic prefix is a repeated part of the OFDM symbol. By analyzing the autocorrelation of the signal, we can identify the length of the cyclic prefix and the spacing between subcarriers.

This research shows that simple methods for synchronization may not always work, especially when only using the cyclic prefix. To improve accuracy, the study removes errors from synchronization by examining phase differences between two consecutive OFDM symbols.

#Building Features for Classification

Once the necessary parameters are estimated, the next step is to create features that clearly represent the different modulation schemes. The features need to be robust against synchronization errors.

In this process, the feature extraction method focuses on capturing the modulation characteristics based on the sampled time-domain sequence. The first goal is to ensure that the sampled sequence is entirely within a single OFDM symbol. The second goal is to eliminate any phase drift caused by synchronization errors.

The features are converted into a 2D histogram that shows the phase and amplitude distribution. This histogram acts as input for a convolutional neural network (CNN)-based classifier.

#The Convolutional Neural Network Classifier

The classifier is designed to process the 2D histograms created during the feature extraction phase. CNNs are a type of neural network that works well for classifying visual data. In this case, they can distinguish between different modulation schemes based on the histogram representation of the features.

The classifier goes through two processing steps before the histogram is inputted. First, the phase differences are adjusted to remove rotations that don’t change the actual signal characteristics. Next, a normalized histogram of the amplitude distribution is created, which helps the classifier understand the modulation better.

#Testing the Classifier

To evaluate how well the proposed classification system works, both synthetic and real-world data are used. The synthetic data is generated under various conditions like different signal-to-noise ratios (SNR). The real-world data is captured using software-defined radios in actual wireless environments.

The performance of the classifier is measured by how accurately it identifies different modulation formats under different noise conditions. It is found that, for both the synthetic and real-world data, the method can reach a minimum accuracy of 97% when the SNR is above the minimum required for reliable communication.

#Results of Classification Accuracy

The results demonstrate that the proposed method for classifying modulation signals works well under various conditions. With synthetic data, the accuracy for estimating parameters and identifying modulation schemes remains high. For real-world measured data, the results are also promising, showing that the system can adapt to different configurations without needing prior knowledge of the transmitted signal.

In particular, it is found that high-order modulation schemes, such as 256QAM and 1024QAM, which are often used in state-of-the-art protocols, can be recognized accurately by the classifier. This is a significant advancement for spectrum sensing applications.

#Conclusion

This study shows that it is possible to develop a robust classifier for identifying the modulation of Wi-Fi 6 and 5G signals without needing any specific prior knowledge of the transmission format. By leveraging techniques like cyclic autocorrelation and using a CNN-based classifier, the method successfully estimates important parameters and classifies modulation schemes effectively.

The integration of advanced machine learning techniques with traditional signal processing methods represents a promising approach for future developments in wireless communication systems. As wireless technology continues to evolve, these techniques will play a crucial role in ensuring efficient use of the radio spectrum and improving the performance of wireless networks.