|
mEEC – A Novel Error Estimation Code with
Multi-Dimensional Feature |
Error Estimation Code (EEC) can
be used to estimate the number of errors in a packet transmitted over a wireless
channel. Typically, such packets do not have a lot of errors, and knowing the
numbers of errors will allow the wireless transmitter to use the most efficient
method to recover the errors, such as transmitting just enough number of parity
bits, instead of retransmitting the entire packet.
EEC
typically works by taking samples
from the packet, then calculating a feature
of the samples and transmit the feature as the code. The receiver can locally
calculate the feature based on the received data packet in the same manner,
which will be different from the transmitted feature if there are errors, and
can use the difference to estimate the number of errors in the packet. The
actual EEC usually consists of multiple features calculated in the same manner.
The
main innovation of mEEC is to introduce a novel
multi-dimensional feature and a color assignment as the code, exploiting the
fact that the error ratios in partial packets are typically small. That is, mEEC divides the
sampled data bits into blocks, then groups multiple blocks into a super-block,
and uses only one number, i.e., the color,
to represent all features of the blocks. The advantage of grouping
is that it introduces useful dependencies among the blocks and allows them to
share the cost of covering low probability events.
The
evaluation shows that mEEC achieves smaller
estimation errors than the state of art on metrics such as the relative mean
squared error, on average by more than 10%-20% depending on the packet sizes,
sometimes as high as over 40%, at the same time having less bias.
This research was supported by my NSF
CAREER grant: CAREER: Addressing Fundamental Challenges for Wireless Coverage
Service in the TV White Space. 1149344.
Publication:
Z. Zhang and P. Kumar, “mEEC: A novel error
estimation code with multi-dimensional feature,” in Proc. of IEEE Infocom,
Atlanta, GA, May 2017. 9 pages. Acceptance rate: 20.93% (292/1395).
Code:
A Matlab
implementation of mEEC can be downloaded here. Please feel free to email me for any
questions related to mEEC.
Slides:
mEEC:
A Novel Error Estimation Code with Multi-Dimensional Feature