Not all nearest neighbors are equal in bounded distance decoding

Bounded distance soft decoders guarantee correct decoding at least up to half the minimum Euclidean distance of a code, ½d₀ . Under practical working conditions, performance degradation as compared to optimal decoding-is dominated by the behavior of the decoder outside the (bounded distance) hyper-spheres whose centers are the codewords. We carefully investigate this issue and reveal an interesting phenomenon: there are three different types of nearest neighbors classified according to their affect on the decision region. Simulation results are presented to demonstrate this phenomena