Conclusions

We presented a novel clustering algorithm that is capable of decoding fast flat-fading time-varying BPSK MIMO channels. This algorithm operates on the received data to which a temporal dimension is added, and it exploits constellation geometry to retrieve symbol threads from it. The only supervised part of the presented clustering method is the final decoding phase.

For moderate numbers of transmitting antennas, the presented method only needs transmission of very few pilot symbols to achieve superior BER performance compared to the tested supervised adaptive algorithm. Its results can further be improved by adding more receiver antennas to the MIMO system.

Future research lines include the extension of this algorithm to communication systems with more clusters, which are encountered if more transmit antennas are used or if the constellation alphabet is extended. To that end other system characteristics could be exploited, apart from the constellation geometry, such as space-time coding and in particular orthogonal block coding, for instance the popular Alamouti coding.