Constellation Mapping Design Using K-means Clustering for Analog Joint Source-Channel Coding Based on Shannon-Kotel’nikov Mappings

Analog joint source channel coding (AJSCC) schemes have gained importance, which have shown to approach the best performance theoretically obtainable when transmitted over AWGN channels. Its simplicity and zero-latency transmission make it more suitable for real-time applications than traditional digital system separation principles. However, the sources and mappings in AJSCC are inherently analog and meanwhile digital communications are standard in wire communications, so we need to digitize some aspects of mapping full-resolution constellations to finite constellations. In this work, we discuss how to digitize the Shannon-Kotel'nikov (S-K) mappings for application in real systems or for interfacing with digital systems. We propose an irregular constellation mapping method and apply it to an existing AJSCC model. The experimental results show that the performance is better than the conventional uniform quadrature amplitude modulation (QAM) constellation mapping method.