Combined source-channel coding with modulation organized vector quantization (MOR-VQ)

In recent years there has been an increasing interest in combined source- and channel-coding schemes, where significant zero-redundancy error protection can be obtained. However complexity in design, requirements for known channels often along with complex system design concepts has until now prevented such systems from common use. This papers demonstrates that significant zero-redundancy error protection can be obtained by assuming the modulation set known when designing the source coder. The single, most important factor that improves performance is the proper ordering of the data, often viewed as a computationally complex problem. A novel approach for the solution of this problem is developed, based on neural network theory. An interesting conclusion is that for such a system, the selection of correct size for the modulation set is not critical, as opposed to what is generally assumed, and that in some cases a larger symbol set even outperforms a smaller symbol set under noisy conditions.