Method of enhanced max-log-a posteriori probability processing
AssigneeHer Majesty the Queen in right of Canada, as represented by the Minister of Industry through the Communications Research Centre
InventorsStewart Crozier, Andrew Hunt, John Lodge
Life of this patent
Estimating the term…
Abstract
The invention comprises an enhancement to max-log-APP processing that significantly reduces performance degradation associated with introducing the "max" approximation into log-APP computations, while still maintaining lower computational complexity associated with max-log-APP processing. This enhancement is achieved by adjusting extrinsic information produced by a max-log-APP process where the magnitude of the extrinsic information is reduced, for example, by multiplying it with a scale factor between 0 and 1.
Claims24
A method of max-log-APP processing of input log-probability quantities, comprising the steps of:
- (a) producing first output log-probability quantities using a max-log-APP process based on a constraint set, each of said first output log-probability quantities corresponding to an input log-probability quantity from the input log-probability quantities;
- (b) computing a first extrinsic information value for each first output log-probability quantity from the first output log-probability quantities, said extrinsic information value computed as the difference between said output log-probability quantity and the corresponding input log-probability quantity;
- (c) for each first extrinsic information value, producing a corresponding second extrinsic information value based on the first extrinsic information values,
- at least one of the second extrinsic information values being smaller in magnitude than the corresponding first extrinsic information value, and
- wherein said second extrinsic information value and the corresponding first extrinsic information value correspond to a same input log-probability quantity;
- and,
- (d) adding each of said second extrinsic information values to the corresponding input log-probability quantity to provide second output log-probability quantities.
A method of max-log-APP processing of input log-probability quantities associated with an error correcting code wherein the error-correcting code is a binary code having one or more even-parity constraints, the method comprising the step of:
- for a set of bits with even-parity, storing extrinsic information in the form of a minimum magnitude value, a second magnitude value, a location index associated with the minimum magnitude value, and a sign bit for each bit in the even-parity set.
A method of max-log-APP processing comprising the step of reducing on average the difference between output log-probability quantities and corresponding input log-probability quantities.
Dependencies are inferred from the claim text.