In this work, we propose a multiple Kronecker product-based space-time (MKronST)-based coding for singlehop point-to-point multiple-input multiple-output (MIMO) systems that employ space-time spreading scheme whose certain number of data streams are spatially spread within the antennas of a subarray while other ones are spread across antennas of different subarrays. Capitalizing on a tensor formalism, the received signals are modelled as a constrained quadrilinear parallel factor (PARAFAC) model, which allows us to derive a semi-supervised iterative receiver for joint symbol and channel estimation. Simulation results are provided to evaluate the performance of the proposed receiver. Our numerical results show that the proposed semi-supervised receiver is capable of jointly estimating the symbol and channel matrices with a superior performance in comparison to competing receivers outlined in the literature.