mEEC: A novel error estimation code with multi-dimensional feature

Error estimation code estimates the bit error ratio of the received data bits with low overhead. It has many applications, especially in estimating the number of errors in a packet transmitted over a wireless link. In this paper, we propose a novel error estimation code, mEEC, that outperforms the existing code by more than 10%-20% depending on the packet sizes, at the same time being less biased. mEEC is mainly based on the idea of grouping multiple blocks of sampled data bits into a super-block, thus creating a multi-dimensional feature. It then compresses these features into a single number, called the color, as the coded bits. Through an intelligent coloring scheme, the blocks in a super-block share the cost of covering low probability events, which allows the decoder to recover the actual feature values from the color even in the presence of error. mEEC also adopts a lightweight redistribution step, which is guided by the solution of an optimization problem and further reduces the estimation errors and bias. We also show that mEEC can be implemented with reasonable storage sizes and low time complexity.