Applying Deep Residual Network to Suppress Optical Coding Access Interference

Optical network has a lot of advantage such as high transmit rate, large bandwidth, immunity to electromagnetic interference, low transmission loss and so on. We adopt Spectral-Amplitude Coding (SAC) of Optical code-division multiple-access (OCDMA) as our system framework. In the paper, we aim at intelligent computations to mitigate Gaussian distribution noises and interference from noisy accumulated multi-carrier transmissions. We proposed the deep learning interference cancellation (DLC) which is composed of convolutional neural network (CNN) and deep residual network (ResNet) concept. The DLC conceptually consists of three phase: Noisy signal extraction, Signal reconstruction and Data decoding. We will compare the hard decision and soft decision method and evaluate the performance over optical coding access receiver. In the simulation, the result shows that RIC with embed DLC will have better performance.