Sparse Bayesian learning (SBL) can be implemented with low complexity based on the approximate message passing (AMP) algorithm. However, it does not work well for a generic measurement matrix, which may cause AMP to diverge. Damped AMP has been used for SBL to alleviate divergence issues at the cost of reducing convergence speed. In this work, we propose a new SBL algorithm based on structured variational inference, leveraging AMP with a unitary transformation. Both single measurement vector and multiple measurement vector problems are investigated. It is shown that, compared to state-of-the-art AMP-based SBL algorithms, the proposed UAMP-SBL is more robust and efficient, leading to remarkably better performance.

An alternative simple derivation for the well-known soft-in soft-out LMMSE (linear minimum mean square error) detector in a turbo system is presented. The derivation leads to a concise representation for the LMMSE detector in terms of the extrinsic means and extrinsic variances of the data symbols, and provides new perspectives for its implementation.

In medium-size massive MIMO systems, the minimum mean-square-error parallel interference cancellation (MMSE-PIC)-based soft-input soft-output (SISO) detector is often used due to its relatively low complexity and good bit error rate (BER) performance. The computational complexity of MMSE-PIC for detecting a block of data is dominated by the computation of a Gram matrix and a matrix inversion. They have computational complexity of O(K ² M) and O(K ³ ), respectively, where K is the number of uplink users with one transmit antenna each and M is the number of receive antennas at the base station. In this letter, by using an L (typically L ≤ 3) terms of Neumann series expansion to approximate the matrix inversion, we reduce the total computational complexity to O(LK M). Compared with alternative algorithms, which focus on reducing the complexity of the matrix inversion only, the proposed method can also avoid calculating the Gram matrix explicitly and thus significantly reducing the total complexity.

Covariance-based detection is a low-complexity blind spectrum sensing scheme that exploits spatial and/or temporal correlations of primary signals. However, its performance severely degrades with the decrease of signal correlations. In this work, a weighted-covariance-based detector is proposed by introducing data-aided weights to the covariance matrix. The false alarm probability, decision threshold, and detection probability are analyzed in the low signal-to-noise ratio (SNR) regime, and their approximate analytical expressions are derived based on the central limit theorem. The analyses are verified through simulations. Experiments with simulated multiple-antenna signals and field measurement digital television signals show that the proposed weighted detection can significantly outperform the original covariance-based detection.

In acoustic event detection, the training data size of some acoustic events are often small and imbalanced. To deal with these, this paper proposes to generate the virtual training data categorically using the auxiliary classifier generative adversarial networks. Soft labels of acoustic events are first calculated to represent the acoustic event localization information. The closer the current frame is to the middle of the manually labeled acoustic event, the higher the soft label will be, which makes the soft labels positively correlated to the acoustic event localization. Then the acoustic event class and the quantized soft labels are used as the input condition to the auxiliary classifier generative adversarial networks to generate arbitrary number of training samples. Experimental results on the TUT Sound Event 2016 under the home environment and TUT Sound Event 2017 under the street environment demonstrate the improved performance of the proposed technique compared to existing acoustic event detection systems.