Improved Detected Data Processing for Decision-Directed Tracking of MIMO Channels

This paper addresses the issue of how to process detected data vectors for decision-directed MIMO channel tracking. We differentiate between a serial processing where the decision-directed estimates are computed after each vector has been detected and a parallel one where a block of detected vectors is processed at once which necessitates a matrix inversion. The probability of existence of a matrix inverse is thoroughly investigated. Parallel processing is improved by deriving the optimal block size. A comparison shows that serial processing performs better in case of fast channel variations.