Yeqing Wu

Cross-Layer Forward Error Correction Scheme Using Raptor and RCPC Codes for Prioritized Video Transmission Over Wireless Channels

The unequal error protection (UEP) has shown promising results for transmitting video over error-prone wireless channels. In this paper, we investigate the cross-layer design of forward error correction (FEC) schemes by using the UEP Raptor codes at the application layer (AL) and UEP rate compatible punctured convolutional (RCPC) codes at physical layer (PHY) for prioritized video packets. The video packets are prioritized based on their contribution to the received video quality. A genetic algorithm (GA)-based optimization algorithm is proposed to find the optimal parameters for both Raptor and RCPC codes, to minimize the video distortion and maximize the peak signal-to-noise-ratio for the given video bit rates and channel constraints (i.e., SNR and available bandwidth). We evaluate the performance of four combinations of the UEP schemes for H.264/AVC encoded video sequences over the AWGN and Rayleigh fading channels and show the superiority of the optimized cross-layer UEP FEC scheme. For Rayleigh fading channel, the proposed cross-layer optimization uses two different time-scales at AL and PHY which allows PHY to adapt faster to the changing channel quality.

Unsupervised Multi-Level Non-Negative Matrix Factorization Model: Binary Data Case

Rank determination issue is one of the most significant issues in non-negative matrix factorization (NMF) research. However, rank determination problem has not received so much emphasis as sparseness regularization problem. Usually, the rank of base matrix needs to be assumed. In this paper, we propose an unsupervised multi-level non-negative matrix factorization model to extract the hidden data structure and seek the rank of base matrix. From machine learning point of view, the learning result depends on its prior knowledge. In our unsupervised multi-level model, we construct a three-level data structure for non-negative matrix factorization algorithm. Such a construction could apply more prior knowledge to the algorithm and obtain a better approximation of real data structure. The final bases selection is achieved through L2-norm optimization. We implement our experiment via binary datasets. The results demonstrate that our approach is able to retrieve the hidden structure of data, thus determine the correct rank of base matrix.

Non-informative hierarchical Bayesian inference for non-negative matrix factorization

Non-negative matrix factorization (NMF) is an intuitive, non-negative, and interpretable approximation method. Canonical NMF approach could derive some basic components to represent original data, while probabilistic NMF approaches try to introduce some reasonable constraints to optimize the canonical NMF model. However, both of them cannot handle ground-truth bases discovering and model order determination problems. In general, the model order of basis matrix needs to be pre-defined. The model order determines the capability and accuracy of data structure discovering. However, how to accurately infer the model order of basis matrix has not been well investigated. In this paper, we propose a method called non-informative hierarchical Bayesian non-negative matrix factorization (NHBNMF) to automatically determine the model order and discover the data structure. They are achieved through hierarchical Bayesian inference model, maximum a posteriori (MAP) criterion, and non-informative parameters. In NHBNMF method, we first introduce a structure with two-level parameters to enable the entire model to approach the distributions of ground-truth bases. Then we use non-informative parameter scheme to eliminate the hyper-parameter to enable automatic searching. Finally, the model order and ground-truth bases are discovered by using MAP criterion and L2-norm selection. The experiments are conducted based on both synthetic and real-world datasets to show the effectiveness of our algorithm. The results demonstrate that our algorithm can accurately estimate the model order and discover the ground-truth bases. Even for the complicated FERET facial dataset, our algorithm still obtained interpretable bases and achieved satisfactory accuracy of the model order estimation. HighlightsA non-informative hierarchical Bayesian non-negative matrix factorization (NHBNMF) algorithm is proposed.The NHBNMF algorithm can automatically find a set of bases which are close to the set of ground-truth bases.Non-informative parameter is employed to enable automatic bases determination.NHBNMF has satisfied performance on several kinds of data sets.

Primate-inspired adaptive routing in intermittently connected mobile communication systems

Abstract An intermittently connected mobile ad hoc network is a special type of wireless mobile network without fully connected path between the source and destination most of the time. In some related works on mobility models, the missing realism of mobility model has been discussed. However, very few routing protocols based on realistic mobility models have been proposed so far. In this paper, we present a primate-inspired mobility model for intermittently connected mobile networks. Such a mobility model can represent and reflect the mobile features of humans. Traditional routing schemes in intermittently connected mobile networks fail to integrate the mobility model with routing strategy to fully utilize the mobility features. To overcome such a drawback, we propose a new routing scheme called primate-inspired adaptive routing protocol (PARP), which can utilize the features of the primate mobility to assist routing. Furthermore, our proposed protocol can determine the number of message copies and the routing strategy based on the walking length of the mobility model. The predictions of the walking lengths are implemented by a particle filter based algorithm. Our results demonstrate that PARP can achieve a better performance than a few typical routing protocols for intermittently connected mobile ad hoc networks.

Spectrum handoffs with mixed-priority queueing model over Cognitive Radio Networks

In this paper, we propose a mixed preemptive and non-preemptive resume priority (PRP/NPRP) M/G/1 queueing model for the modeling of the traffic in Cognitive Radio (CR) Networks with prioritized transmissions. A traffic-adaptive spectrum handoff scheme is then developed based on the proposed queueing model for delay sensitive applications. This spectrum handoff scheme reduces the delivery time of the delay-sensitive applications for secondary users (SUs) while ensuring the overall performance of the network by avoiding overly frequent spectrum handoff between SUs. In the experiment section, we compare with two recently proposed queueing models using the traffic-adaptive spectrum handoff scheme and show the superior performance of the proposed approach.

Optimal Spectrum Handoff Control for CRN Based on Hybrid Priority Queuing and Multi-Teacher Apprentice Learning

An optimal spectrum handoff scheme for cognitive radio networks (CRNs) is presented in this paper. This scheme has two novel features: 1) Hybrid rule-based priority queuing model: To overcome the limitations of preemptive resume priority and nonpreemptive resume priority (PRP/NPRP) queuing models, a hybrid queuing model with discretion rule is proposed to characterize the spectrum access priority among secondary users (SUs). This hybrid queuing model is then used to calculate the channel waiting time during spectrum handoff; and 2) Multiteacher apprentice learning: Unlike existing CRN cognition engine designs that focus on spectrum adaptation through SU self-learning (i.e., an SU learns how to adapt to the dynamic CRN environment by itself), we propose the concept of multiteacher knowledge transfer, wherein the multiple SUs that already have mature spectrum adaptation strategies share their knowledge with an inexperienced SU. Our simulation results show that the proposed new designs improve the spectrum handoff accuracy for the complex CRN environments.

Apprenticeship Learning Based Spectrum Decision in Multi-Channel Wireless Mesh Networks with Multi-Beam Antennas

We propose a novel spectrum decision scheme (i.e., channel selection and handoff) for wireless mesh networks (WMN) which use multiple channels and nodes equipped with multi-beam directional antennas. Our scheme has the following features: (i) It performs spectrum decision by considering various WMN parameters, including the channel quality, beam orientation, antenna-caused deafness and capture effects, and application priority level. (ii) It uses the reinforcement learning (RL)-based spectrum decision process to achieve the optimal quality of multimedia transmission in the long term. However, a newly-joined WMN node could take a long time to make a correct spectrum decision due to the difficult choice of initial RL parameters. Therefore, our scheme uses the apprenticeship learning in conjunction with the RL model, to speed up the spectrum decision process by choosing a suitable neighboring node (called “expert”) to teach a newly-joined node (called “apprentice”). Our experiments demonstrate that the proposed spectrum decision scheme improves the network performance and multimedia transmission quality.

Feature-based compressive signal processing (CSP) measurement design for the pattern analysis of Cognitive Radio spectrum

Cognitive Radio (CR) can efficiently utilize the licensed wideband spectrum whenever the primary users (PUs) are absent. Spectrum sensing is the first step and an important function to fulfill the CR system. A cyclostationary feature detector can robustly detect the PUs modulated signals even under strong interferences. However this detector needs high signal sampling rate and also puts heavy computation burden on the system. Compressive sensing (CS) can compress the data at the front sampling end but has high overload and delay from the reconstruction side. In this work we generate the compressive CR spectrum measurement by utilizing both the cyclostationary feature and sparsity prior knowledge at the spectrum sensing front end, and we apply the compressive signal processing (CSP) without the need of signal or feature reconstruction. This can significantly shorten the CR spectrum sensing time. Our experimental results have shown the pattern analysis accuracy and efficiency of our CSP scheme.

Measuring activities and counting steps with the SmartSocks - An unobtrusive and accurate method

Physical inactivity is an important contributor to non-communicable diseases in countries of high income, and increasingly so in those of low and middle income. Physical inactivity is the leading cause of many diseases. It has been estimated that as many as 250,000 deaths per year in the United States, approximately 12% of the total, are attributable to a lack of regular physical activity. Measuring physical activities and counting steps is an effective method to diagnose some diseases. It can also serve as an effective method to encourage people to increase their physical activity. Pedometers have been invented as a convenient way of counting steps. However most of them lack the functionality of differentiating activities. Pressure sensor pads can measure steps and gait, but as the pad has a limited size, it can not meet the need of anytime and anywhere usage. In this study, we made the Sensor Socks for measuring physical activities and counting steps. It is unobtrusive and convenient for everyday usage. Our experimental results show that the system has a high accuracy of the classification of physical activities and counting steps in a home or community environment.

A Fast Raptor Codes Decoding Strategy for Real-Time Communication Systems

We propose an efficient algorithm for Raptor decoding, which reduces the computational complexity of the most time-consuming steps in systematic decoding. Our proposed algorithm includes two aspects: First, to handle the decoding failure of the Raptor decoding, we propose a scheme, which is called the No-Wrapup Failure Handling scheme. It can resume the decoding process from where it fails after receiving a pre-defined number of additional encoded symbols, and thus avoids the repetition of time-consuming steps in the decoding process. Second, in order to reduce the time of finding the row with the minimum degree in the precode, we propose a Fast Min-Degree Seeking (FMDS) scheme. FMDS automatically maintains and updates the row degrees of the precode when converting the precode into an identity matrix through Gaussian elimination and Belief-propagation. Experimental results show that, compared to other Raptor decoding schemes, the proposed scheme achieves a much shorter decoding time, and can greatly speed up the data recovery in real-time applications.