Modulated Autocorrelation Convolution Networks for Automatic Modulation Classification Based on Small Sample Set

Abstract

For modulation classification, hand-crafted approaches can generalize well from a few samples, yet deep learning algorithms require millions of samples to achieve the superior performance with purely data-driven manner. However for many practical problems only with small sample set (SSS) available, there still remains a challenge for deep learning. In this paper, we employ deep learning to solve the modulation classification task in a more practical setting, particularly suffering from the SSS problem and with low signal-to-noise ratios (SNRs). Novel modulated autocorrelation convolution networks (MACNs) are introduced to capture periodic representation for automatic modulation classification (AMC). In MACNs, modulated communication signals are classified with the periodic local features under an autocorrelation convolution criterion. Modulation filters are utilized to enhance the capacity of the convolution filters and compress the model. On a challenging SSS learning task in low SNRs, MACNs achieve state-of-the-art performance that outperforms the existing algorithms for AMC, while compressing the size of required storage space of convolutional filters by a factor of 8 compared with convolution neural networks (CNNs).

Object Detection Recognition and Robot Grasping Based on Machine Learning: A Survey

“…It is found that the functions of different algorithms are varied, indicating that each algorithm has different application scenarios. Although deep learning plays a dominant role in the field of machine vision, deep learning is data-driven and has poor performance in small datasets [52][53][54]. However, traditional machine learning can adapt to a variety of datasets; especially in scenarios with small amounts of data (such as the medical field), machine learning has better performance [55,56].…”

Deep-Learning-Based Classification of Digitally Modulated Signals Using Capsule Networks and Cyclic Cumulants

“…As an alternative to the conventional methods mentioned above, in recent years, deep-learning (DL)-based techniques employing neural networks (NNs) have been explored for classifying signals using specific features that can be extracted from the data [ 17 , 18 , 19 , 20 , 21 , 22 ]. We note that NNs require extensive training to become proficient and that, depending on the type of NNs and features used, overfitting and a lack of generalization can affect the robustness of the trained NNs such that their performance degrades when presented with inputs that have different probability distributions than those in the training dataset.…”