Speech signal degradation in real environments mainly results from room reverberation and concurrent noise. While human listening is robust in complex auditory scenes, current speech segregation algorithms do not perform well in noisy and reverberant environments. We treat the binaural segregation problem as binary classification, and employ deep neural networks (DNNs) for the classification task. The binaural features of the interaural time difference and interaural level difference are used as the main auditory features for classification. The monaural feature of gammatone frequency cepstral coefficients is also used to improve classification performance, especially when interference and target speech are collocated or very close to one another. We systematically examine DNN generalization to untrained spatial configurations. Evaluations and comparisons show that DNN-based binaural classification produces superior segregation performance in a variety of multisource and reverberant conditions.

In this letter, a side-peak cancellation unambiguous acquisition technique is proposed for sine-phased binary offset carrier (BOC) modulated signals. The test criterion used in this technique is based on a synthesized correlation function which has no major positive side peak. This synthesized correlation function is obtained by subtracting the cross-correlation between the received sin-BOC signal and an auxiliary signal from the autocorrelation of sin-BOC signal. For different types of BOC signal, the proposed technique employs corresponding modulated symbols of the auxiliary signal. The common solution of the symbol shape vector is derived, and the theoretical false alarm and detection performance formulas are given. Theoretical and simulation results show that at the expense of some performance degradation this technique completely removes the ambiguity threat in acquisition process.

In this letter, we propose an unambiguous tracking technique for the new multiplexed binary offset carrier (MBOC) modulated signals, which will most likely be employed in both European Galileo system and modernized global positioning system (GPS). The discriminator used in this technique is based on a pseudo correlation function. It uses two kinds of gating correlators and a novel combination function, which completely removes side peaks from the correlation function while keeping the sharp main peak. Results demonstrate that this technique is totally unambiguous while maintaining the same level of tracking performance with respect to thermal noise as the traditional MBOC tracking method.

Because of the long period and high chip rate of GPS P-code, direct acquisition is challenging. In the letter, the widely used zero-padding scheme (ZPS) for direct GPS P-code acquisition is generalized to investigate the effects of the ZPS on detection performance, parallel searching capability, and mean acquisition time. It is shown that, by adjusting the zeros padded to received signal according to signal to noise ratio, the generalized zero-padding scheme (GZPS) makes a better tradeoff between detection performance and parallel searching capability and further reduces mean acquisition time. The generalized zero-padding scheme can be easily applied to the previously proposed zero-padding method (ZPM) and the direct average method (DAM) to improve their mean acquisition time performance and it does not increase the implementation complexity of a receiver.

For long PN-code, rapid acquisition is difficult due to large search space. To speed up the search process, extended replica folding acquisition search technique (XFAST), which directly reduces the code phases to be searched by folding local signal, provides an efficient approach to rapid acquisition. Nevertheless, after folding the correlation properties of PNcode are degraded; hence, the detection performance of XFAST to weak signal is worse than that of nonfolding methods. To improve the detection performance, a dual-folding acquisition method (DF) is proposed. By folding both incoming signal and local signal, DF extends coherent integration time to enhance detection performance and indirectly reduce mean acquisition time. Numerical results demonstrate the enhancement of the proposed method with respect to other methods such as serial search (SS), zero-padding method (ZP), and XFAST.

Due to the functional complexity of its signal-to-interference (SIR) ratio, the outage probability of maximum ratio combining is available only for certain independent and identically distributed (i.i.d.) fading environments. In this paper, we partially relax this restriction by allowing the signal channel-gain vector to follow a general fading distribution with arbitrary spatial correlation. The problem is tackled in a novel framework. The key step is to represent the probability density of the reciprocal of SIR, conditioned on signal vector, as the higher-order derivative of a simple exponential function in signal power whereby a generic formula for outage probability can be determined. The application of the generic formula to Rayleigh, Rician, and Nakagami faded signals is elaborated. Numerical results are also presented for illustration.

No systematic procedure for tightly bounding the average capacity of multiple-input-multiple-output (MIMO) correlated Rician fading channels is available in the literature. In addition to the involvement of a highly nonlinear log-determinant operator in the conditional capacity expression, the difficulty arises from the complicated noncentral Wishart distribution of channel sample matrix. In this paper, we tackle the problem with arbitrary antenna correlation existing either at the transmitter or at the receiver, but allowing for the numbers of the transmit and receive antennas to be arbitrary. By introducing an exact determinant expansion and by finding an explicit expression for the general moment of the determinant of the channel sample matrix, we obtain a general upper bound for the average channel capacity. To obtain a general lower bound, we construct and prove a multivariate convex function with each of its variables being the log-determinant function of a complex noncentral Wishart-distributed matrix. We further show that the general bounds so obtained can be simplified to explicit expressions for Rician fading channels with arbitrary semicorrelation and a mean matrix of rank one. The new results are simple, easy to be used, and superior in tightness as evidenced by intensive numerical examples.

No systematic procedure for tightly bounding the average capacity of MIMO correlated Rician fading channels is available in the literature. The difficulty arises from the complicated non-central Wishart distribution of channel sample matrix, in addition to the involvement of a highly nonlinear log determinant operator in the conditional capacity expression. In this paper, we tackle the problem with arbitrary antenna correlation existing either at the transmitter or at the receiver, but allowing for the numbers of the transmit and receive antennas to be arbitrary. By introducing an exact determinant expansion and by finding the explicit expressions for the expected values of the determinant and log-determinant of the complex non-central Wishart-distributed matrix, we obtain a simple but very tight upper and lower bounds for the average channel capacity. Both bounds are simple, in closed form, easy to be used, and superior in tightness as evidenced by intensive numerical examples.

In this paper, we propose a new algorithm for active queue management, called ERB, which is based not only on the queue size, but also on the estimation of the rate and burstiness of the aggregate enqueueing traffic. By adjusting the packet marking probability according to these measured and estimated information in the queue management, ERB can make the router absorb the burstiness and adapt to various network conditions and traffic patterns intelligently. As a result, ERB derives the adaptability and robustness in the complex network environment. Simulation results confirm the validity of this algorithm and compare the performance of ERB, DropTail, RED and REM algorithms.

A new fairness criterion (logarithmic proportional fairness) is introduced in this paper. Under the control of this criterion, the tradeoff among fairness, network utility and total delay can be achieved in the Internet. This paper also presents a new scheduling algorithm-logarithmic proportional fair queueing scheduling algorithm (LPFQ, in short). By using this scheduling algorithm in routers, the logarithmic proportional fairness criterion and punishment to misbehaving flows can be implemented in the Internet. The theoretical analysis and simulation results prove their validity.