Emotion recognition based on Electroencephalogram (EEG) is of great importance for cognitive psychology, disease therapy, etc. However, most mainstream recognition models are trained using batch learning, which fails to adapt to the dynamic and non-stationary nature of real-time EEG streams. In contrast, online learning can adjust model parameters continuously, but it typically relies on labeled data and a fully trained initial model, which are unavailable in practical scenarios. To tackle this challenge, we propose a novel learn-review-challenge online meta-learning framework (LRCOMF) for unlabeled online EEG learning. This framework incorporates a meta updating module through a multi-task cache and a customized sampling strategy to improve the model's generalization during online learning. A sample judgement module is implemented based on a prototype weight being designed to estimate the confidence of predicted labels during the "learn-review" step. Additionally, a challenge module using a clustering quality metric determines whether low-confidence samples can be reconsidered during the "challenge" phase. The validation employed the DEAP and DREAMER datasets. In comparison to the initial baseline condition, the recognition model exhibited substantial enhancements 3.59%/3.87%/1.98% (p $< $ 0.001) for arousal, valence, dominance labels in the DEAP dataset. Substantial increases of 3.42%/3.63%/2.93% (p $< $ 0.001) were also noted in DREAMER's evaluation of arousal/valence/dominance labels. This research is essential for tackling unlabeled online learning in authentic EEG contexts and enhancing the advancement of EEG practical applications.

In the rapidly evolving field of video traffic, ensuring a smooth video streaming experience for users is critical for network operators. Accurately and promptly detecting stalling events, a significant indicator of poor quality of experience, remains challenging due to varying detection time resolutions in existing techniques, which often detect stalls every video chunk, or every five or ten seconds. This paper makes three key contributions. First, we introduce the concept of detection granularities to enable fair performance comparisons and reveal their impact on detection performance from the data sampling perspective. Second, we propose a novel feature extraction approach that captures both packet-level and chunk-level features in a unified sequential manner to effectively detect stalling events. Third, a novel sample reweighting method is proposed to address the detection timeliness problem by focusing more on difficult samples around stalling starting or ending. Experimental results on both video-on-demand and live streaming traces demonstrate that our feature extraction approach achieves an average improvement of 5.3% in f1-score, 4.7% in coverage rate, and reduces stalling response time by 0.4 seconds compared to existing techniques. Additionally, the sample reweighting method further improves the detection sensitivity without compromising f1-scores for all detection techniques.

To improve both sensing and communication performances, this paper proposes a coordinated multi-point (CoMP) transmission design for a dual-functional radar-communication (DFRC) system. In the proposed CoMP-DFRC system, the central processor (CP) coordinates multiple base stations (BSs) to transmit both the communication signal and the dedicated probing signal. The communication performance and the sensing performance are both evaluated by the signal-to-interference-plus-noise ratio (SINR). Given the limited backhaul capacity, we study the waveform and clustering design from both the radar-centric perspective and the communication-centric perspective. Dinkelbach's transform is adopted to handle the single-ratio fractional objective for the radar-centric problem. For the communication-centric problem, we adopt quadratic transform to convexitify the multi-ratio fractional objective. Then, the rank-one constraint of communication beamforming vector is relaxed by semidefinite relaxation (SDR), and the tightness of SDR is further proved to guarantee the optimal waveform design with fixed clustering. For dynamic clustering, equivalent continuous functions are used to represent the non-continuous clustering variables. Successive convex approximation (SCA) is further utilized to convexitify the equivalent functions. Simulation results are provided to verify the effectiveness of all proposed designs.

Keyword Spotting (KWS) plays a vital role in human-computer interaction for smart on-device terminals and service robots.It remains challenging to achieve the trade-off between small footprint and high accuracy for KWS task.In this paper, we explore the application of multi-scale temporal modeling to the smallfootprint keyword spotting task.We propose a multi-branch temporal convolution module (MTConv), a CNN block consisting of multiple temporal convolution filters with different kernel sizes, which enriches temporal feature space.Besides, taking advantage of temporal and depthwise convolution, a temporal efficient neural network (TENet) is designed for KWS system 1 .Based on the purposed model, we replace standard temporal convolution layers with MTConvs that can be trained for better performance.While at the inference stage, the MTConv can be equivalently converted to the base convolution architecture, so that no extra parameters and computational costs are added compared to the base model.The results on Google Speech Command Dataset show that one of our models trained with MTConv performs the accuracy of 96.8% with only 100K parameters.

The next generation wireless network is expected to connect billions of nodes, which brings up the bottleneck on the communication speed for distributed data fusion. To overcome this challenge, computation over multiple access channel (CoMAC) was recently developed to compute the desired functions with a summation structure (e.g., mean, norm, etc.) by using the superposition property of wireless channels. This work aims to maximize the achievable function rate of reliable CoMAC in wireless networks. More specifically, considering channel fading and transceiver design, we derive the achievable function rate adopting the quantization and the nested lattice coding, which is determined by the number of nodes, the maximum value of messages and the quantization error threshold. Based on the derived result, the transceiver design is optimized to maximize the achievable function rate of the network. We first study a single cluster network without inter-cluster interference (ICI). Then, a multi-cluster network is further analyzed in which the clusters work in the same channel with ICI. In order to avoid the global channel state information (CSI) aggregation during the optimization, a low-complexity signaling procedure irrelevant with the number of nodes is proposed utilizing the channel reciprocity and the defined effective CSI.

End-to-end encryption challenges mobile network operators to assess the quality of the HTTP Adaptive Streaming (HAS), where the quality assessment is coarse-grained, e.g., detecting if there exist stalling during the whole playback. Targeting on this issue, this paper proposes an attention-based hybrid RNN-HMM model, which integrates HMM with attention mechanism to predict the player states. The model is trained and evaluated based on the download speed and player state sequences of encrypted video sessions collected from YouTube. Experiment results show that the proposed model is able to recognize the player states with 86.53% ~ 94.35% accuracy, and thus achieves to assess video quality in a fine-grained manner, where how long the stalling lasts and when the stalling occurs can be evaluated effectively from the download speed sequence even when encryption is employed.

Hyperspectral image (HSI) denoising plays an important role to enhance the image quality for subsequent applications. This paper proposes a novel denoising framework for the HSI, in which the denoising issue is modeled as a convolutional neural network (CNN) constrained non-negative matrix factorization problem. Then, by adopting the proximal alternating linearized minimization, the proposed approach can be decomposed into two iterative steps: In the first step, the spectral matrix is updated using a proximal operator to guarantee the non-negativity; in the second step, the abundance matrix is updated using a designed CNN. Finally, we propose a transfer learning scheme to train the design CNN with a natural gray image dataset. Exhaustive experiments show that the proposed approach outperforms the comparison state-of-the-art methods on two different test HSI datasets.

In recent years, millimeter wave (mmWave) communications, considered as a compelling technology to achieve throughput improvement target of the fifth generation (5G) mobile networks, is quickly attracting attention in both scientific and industrial communities. To enhance reliability of mmWave link, mmWave-LTE networks with MPTCP could be used as an effective solution due to its path diversity. However, the poor performance of MPTCP is found in mmWave-LTE networks based on previous studies. In this paper, we identify the reason of the preceding problem which is the congestion window overgrowth in Slow Start phase, and propose a BDP Estimation Based Slow Start (BESS) algorithm for MPTCP to address this problem. Our simulations show that with the control of BESS, MPTCP can sufficiently use the bandwidth of mmWave networks, and improve networks robustness when mmWave link is switched from LOS to NLOS condition.

Multipath transmission control protocol (MPTCP) allows TCP (transmission control protocol) connections to operate across multiple paths concurrently to help improve resource utilization as well as connection robustness. Existing congestion algorithms for MPTCP are mainly loss-based, thus leading to severe performance degradation in wireless environment and buffer bloat on network devices. In this paper, to get rid of these shortcomings, we modify a congestion based TCP congestion control algorithm (bottleneck bandwidth and round-trip propagation time, or BBR) and propose a new congestion control algorithm for MPTCP based on bottleneck estimation, whose name is weighted BBR (wBBR). We introduce weight factors to subflows in MPTCP flows to control convergence bandwidth of each subflow, and then propose an iterative algorithm to lead the network to a fairer convergence when BBR/wBBR flows sharing bottleneck links. Simulation results show that our design has better performance in terms of bottleneck fairness compared with other MPTCP congestion control algorithms. And of course, it inherits the advantages of the BBR algorithm - let the network work at the point with max bandwidth and min queue.

The mobility models obtained from mobile data are expected to affect numbers of fields, including urban planning, road traffic engineering, human sociology, epidemiology of infectious diseases, or telecommunication networks. Current user mobility models are mainly extracted from Call Detail Records (CDR) data or WiFi traces. However, CDR data only captures user movements during telephone calls or short message service (SMS) and WiFi traces can not provide intuitive understanding to mobility behavior of cellular network users in large scale. In this paper, we take the first step to investigate if the characteristics of mobility derived from 4G cellular data network is different from the previous findings utilizing other data sources, especially the widely used, CDR based approach. Utilizing our Hadoop based mobile big data processing platform together with a systematic analysis framework of user mobility, we present a comprehensive characterization of the mobility models from the 6 Terabyte (TB) 4G data traffic. Through the comparison with CDR models and 3G models from the view of user occurrence patterns, user movement patterns and dominant locations of users, we find that the 4G data traffic can provide finer granularity of mobility and location information.

With the exponential growth of digital cameras and the aid of powerful video editing softwares, videos with various resolutions become ever more popular. Therefore, there is a great demand for video coding schemes supporting arbitrary resolutions with higher efficiency. In the latest standard, only a simple padding with direct copying is adopted to meet this requirement, and the coding efficiency is still unsatisfactory. To improve the performance, a novel two-stage picture padding is proposed in this paper. Based on the adopted transform, an efficient residual padding is employed in the first stage to minimize the burdened coding bits of padded parts. To obtain a reference picture with high-quality boundary, a template matching based approach is also proposed to rearrange the padded pixels in the second stage. Experimental results reveal that on top of HEVC, our method offers the performance with 2.1%, 1.8%, and 1.5% BD-rate reduction for RA-Main, LDB-Main and LDP-Main configurations, respectively. Compared with the state-of-the-art algorithm, it still outperforms for kinds of test sequences under all configurations.

In this paper, a novel cell cooperation-based scheme is proposed as a means of energy saving in an overlapped coverage area of several cells. Different from the existing singlecell serving mode, our proposal utilizes virtual cell clustering (VCC) technique, which aggregates spare bandwidth from all adjacent cells, to reduce energy consumption while satisfying the service requirement simultaneously. For the maximum energy saving gain, a service partition procedure is then introduced with an optimization problem modeled in the scenarios under consideration. Evidences from numerical simulations further demonstrate that the proposed scheme outperforms the existing mode with a significant energy saving achievement.

In this paper, we investigate the utility-based resource allocation problem in OFDMA relay networks with service differentiation, where Rate Constraint (RC) and Best Effort (BE) services are supported. Our objective of resource allocation is to maximize the sum utility of BE service users, while guaranteeing that the utility of each RC service user is equal to one. A joint optimization problem for relay selection, subcarrier assignment and power allocation is formulated. Since the problem cannot be solved directly, we make continuous relaxation and solve it by Lagrangian dual method. The optimal allocation strategy is obtained using Karush-Kuhn-Tucker (KKT) conditions. As it is difficult to converge to the optimal solution, thus we further present a heuristic resource allocation algorithm with low complexity. Simulation results show that our proposed algorithm achieves higher utility of BE users and lower outage probability of RC users, and also yields a good tradeoff between system throughput and user fairness.

A distributed scheduling algorithm is proposed for achieving end-to-end flow max-min fairness in wireless multi-hop networks. The problem is modeled as an optimization problem by taking into account the MAC-layer constraints. With dual programming and Lagrange relaxation, the problem is decomposed into two subproblems, one is for transport layer at the source node, and the other for MAC layer in the network. A max-price based max-min fair rate allocation algorithm is adopted for the transport layer subproblem and an algorithm based on integer programming is designed for scheduling link activities at MAC layer. Simulation results show that the proposed algorithm can achieve conflict-free schedule for wireless links and provide max-min fairness for end-to-end flows efficiently.