MIMO Demodulation Complexity Reduction via SINR-Based Bit Segmentation

The patent segments the demodulation process into two parts: first determining SINR for each path and selecting reliable bits based on SINR thresholds, then applying ML algorithm only to these reliable bits. This segmentation reduces the input size for the complex ML algorithm while maintaining high data rates through higher-order modulation.

The patent applies the ML algorithm partially - only to bits that are determined to be reliable based on SINR criteria, rather than processing all bits through the full ML demodulation process. This partial application significantly reduces computational complexity while maintaining performance for the most important bits.

The patent applies different processing quality to different bits: reliable bits (those meeting SINR thresholds) undergo full ML demodulation for high accuracy, while unreliable bits are handled differently. This local differentiation maintains demodulation accuracy for critical bits while reducing overall computational burden.

The patent performs partial ML demodulation - applying the computationally intensive ML algorithm only to a subset of reliable bits rather than all bits. This partial action maintains sufficient demodulation accuracy for the most important information while making the computational burden acceptable.