RQ-VAE-based semantic comunication for image transmission in an N-D CAP underwater wireless optical communication system

Underwater wireless optical communication (UWOC) has been regarded as a promising technology for providing high-speed underwater data transmission. Traditional UWOC systems primarily focus on the accurate transmission of bit sequences, aiming to minimize bit error rates through sophisticated source and channel coding. However, this approach often leads to inefficiencies in challenging environments. Semantic communication emerges as a novel communication paradigm that shifts the objective from exact bit recovery to the faithful reconstruction of semantic information behind source data, enhancing robustness and efficiency. In this paper, we propose a semantic communication framework based on residual quantization variational autoencoder (RQ-VAE) with N-dimensional carrier-less amplitude phase (N-D CAP) modulation for image transmission. At the transmitter, the input images are encoded into latent feature vectors, which capture core semantic information. Then, the RQ technique is implemented to approximate these vectors by recursively finding the closest matching codeword in a learned codebook. Instead of directly transmitting the latent feature vectors, we only transmit a stack of multi-stage quantized codes, thereby substantially reducing the data volume. Integrating with N-D CAP, the quantized codes from different stages are in line with the multi-user communication concept of N-D CAP, enabling concurrent transmission over the same frequency band. And we demonstrate this scheme in a 450 nm laser-based UWOC system with a 5 m transmission distance. We adopt the traditional JPEG and LDPC standards as the baseline. Experimental results show that the proposed scheme outperforms the traditional scheme by improving PSNR by up to 4.68 dB at similar frame rates, or increasing frame rate by 329% at similar PSNR. A transmission rate of 4,394,531 image instances/patches per second at 64×64 was achieved. To the best of our knowledge, this is the highest reported rate for underwater image transmission.