Abstract: A method is disclosed for signal detection in a wireless communication system that includes multiple-antenna arrays for transmission and reception. The method involves multiplying a vector of signals collected from respective receiving antennas by a nulling vector, such that the resulting vector product provides an estimated value corresponding to a specific one of the transmitted signals. A respective such multiplication is performed to detect each of the transmitted signals. In contrast to related methods of the prior art, each nulling vector is obtained from a channel matrix of estimated channel coefficients without performing any matrix inversion operations, and only a single pseudoinverse related to the channel matrix is computed. As a result, numerical stability is improved and the computational complexity of the method is reduced by an order of magnitude relative to methods of the prior art.

Abstract: Differential unitary space-time (DUST) communication is a technique for communicating via one or more antennas by encoding the transmitted data differentially using unitary matrices at the transmitter, and decoding differentially without knowing the channel coefficients at the receiver. Since channel knowledge is not required at the receiver, DUST is ideal for use on wireless links where channel tracking is undesirable or infeasible. Disclosed are a class of Cayley codes for DUST communication that can produce sets of unitary matrices that work with any number of antennas, and has efficient encoding and decoding at any rate. The codes are named for their generation via the Cayley transform, which maps the highly nonlinear Stiefel manifold of unitary matrices to the linear space of skew-Hermitian matrices. The Cayley codes can be decoded using either successive nulling/cancelling or sphere decoding.

Abstract: In a method of wireless transmission, space time matrices are used to spread the transmission of data over two or more transmit antennas and/or over two or more symbol intervals. Initially, blocks of data are encoded as symbols, each being a complex amplitude selected from a symbol constellation. A finite set of space-time matrices, referred to as “dispersion matrices,” is predetermined. In transmission, a group of symbols are transmitted concurrently. Each of the symbols to be transmitted is multiplied by a respective dispersion matrix. Thus, a composite matrix, proportional to a sum of dispersion matrices multiplied by their corresponding symbols, is modulated onto a carrier and transmitted. In reception, knowledge of the dispersion matrices is used to recover the transmitted symbols from the received signals corresponding to the composite matrix that was transmitted.

Abstract: A method and apparatus for designing a multilayer feed forward neural network that produces a design having a minimum number of connecting weights is based on a novel iterative procedure for inverting the full Hessian matrix of the neural network. The inversion of the full Hessian matrix results in a practical strategy for pruning weights of a trained neural network. The error caused by pruning is minimized by a correction that is applied to remaining (un-pruned) weights thus reducing the need for retraining. However, retraining may be applied to the network possibly leading to further the simplification of the network design.