Intelligent adaptive coherent optical receiver based on convolutional neural network and clustering algorithm.

Area of Science:

• Optical communication networks
• Signal processing
• Machine learning applications

Background:

• Accurate identification of physical layer information, specifically modulation formats, is crucial in cognitive optical networks.
• Modulation format identification is essential for coherent optical receiver performance, including carrier-phase recovery.
• Constellation diagrams are vulnerable to noise, necessitating robust identification methods.

Purpose of the Study:

• To develop a simple yet effective receiver system for identifying modulation formats in optical signals.
• To enhance the accuracy of symbol decisions by tracking fast signal changes.
• To improve bit error rate (BER) performance in optical communication systems.

Main Methods:

• Utilizing a convolutional neural network (CNN) to process amplitude data post-modulation-format-agnostic clock recovery.
• Employing a density-based clustering algorithm (fast search and find the density peaks) for constellation cluster classification.
• Applying the k-nearest-neighbor (KNN) method for sample labeling.
• Experimental demonstration of the proposed receiver system.

Main Results:

• The proposed system successfully identifies modulation formats based on amplitude information.
• The system demonstrates the ability to track rapid signal changes.
• Significant improvement in bit error rate (BER) was experimentally achieved.

Conclusions:

• The developed receiver architecture offers a straightforward approach to modulation format identification.
• The integration of CNN and clustering methods enhances signal processing capabilities in optical networks.
• The system's performance in tracking signal dynamics leads to improved symbol decision accuracy and overall system performance.