Facial expression recognition (FER), aiming to classify the expression present in the facial image or video, has attracted a lot of research interests in the field of artificial intelligence and multimedia. In terms of video based FER task, it is sensible to capture the dynamic expression variation among the frames to recognize facial expression. However, existing methods directly utilize CNN-RNN or 3D CNN to extract the spatial-temporal features from different facial units, instead of concentrating on a certain region during expression variation capturing, which leads to limited performance in FER. In our paper, we introduce a Graph Convolutional Network (GCN) layer into a common CNN-RNN based model for video-based FER. First, the GCN layer is utilized to learn more significant facial expression features which concentrate on certain regions after sharing information between extracted CNN features of nodes. Then, a LSTM layer is applied to learn long-term dependencies among the GCN learned features to model the variation. In addition, a weight assignment mechanism is also designed to weight the output of different nodes for final classification by characterizing the expression intensities in each frame. To the best of our knowledge, it is the first time to use GCN in FER task. We evaluate our method on three widely-used datasets, CK+, Oulu-CASIA and MMI, and also one challenging wild dataset AFEW8.0, and the experimental results demonstrate that our method has superior performance to existing methods.

Audio adversarial examples, imperceptible to humans, have been constructed to attack automatic speech recognition (ASR) systems. However, the adversarial examples generated by existing approaches usually incorporate noticeable noises, especially during the periods of silences and pauses. Moreover, the added noises often break temporal dependency property of the original audio, which can be easily detected by state-of-the-art defense mechanisms. In this paper, we propose a new Iterative Proportional Clipping (IPC) algorithm that preserves temporal dependency in audios for generating more robust adversarial examples. We are motivated by an observation that the temporal dependency in audios imposes a significant effect on human perception. Following our observation, we leverage a proportional clipping strategy to reduce noise during the low-intensity periods. Experimental results and user study both suggest that the generated adversarial examples can significantly reduce human-perceptible noises and resist the defenses based on the temporal structure.

Nowadays, it is well-known that the LTE (Long Term Evolution) networks have greatly tackled the Doppler effect problem at the physical layer since they are capable of achieving a 100 Mbps data throughput on a high-speed moving vehicle up to 350 km/h within the same cell (i.e. no handover is allowed). However, there are still cases where vehicles are moving at very high speeds and thus frequent handovers across cells are inevitable, such as high-speed trains, which have been constructed rapidly all over the world in recent years. Therefore, how to maintain good link quality and schedule multimedia services optimally on high-speed trains remains challenging. In this paper, we propose a novel optimal LTE-based multimedia-service-scheduling and resource-allocation mechanism for highspeed trains, which could maximize the service rate and maintain a good service quality at the same time. This new scheme makes use of the seamless handover mechanism by carefully organizing the cell array along the railroad and aggregating the data (at the femtocell) from different users within the train cabins. Besides, we also project to schedule the multimedia services and allocate the network resources as fair as possible for all the users on the train. The simulation results justify the effectiveness of our proposed new scheme.

In this letter, we propose new frequency-domain pilot-multiplexing techniques (FDPMTs) for the channel estimation and equalization in orthogonal frequency-division multiplexing (OFDM) systems. A robust and effective pilot insertion and detection scheme is devised thereby. These pilot positions are optimally selected to minimize the distortion of the transmitted time-domain signal caused by the subcarrier-removal at the corresponding pilot positions. Besides, three different blind pilot-detection techniques are designed at the receiver without any a priori knowledge of the pilot positions, and the distorted data symbols can thus be iteratively reconstructed. Rigorous theoretical analysis and Monte Carlo simulation results both demonstrate that our proposed new OFDM system using dynamical pilot positions is more robust than the conventional OFDM system using the fixed pilot positions over multipath fading channels.

Orthogonal frequency-division multiplexing (OFDM) has been widely used for many communication technologies nowadays due to its high spectral efficiency, robustness against inter-symbol-interference and multipath fading, simplicity for implementation, and so on. In this paper, we propose the frequency-domain pilot multiplexing techniques (FDPMTs) for the channel estimation and equalization in OFDM systems. A robust and effective pilot insertion and detection scheme is devised thereby. The information signal sequence resulting from the constellation mapper is spread over all subcarriers by a precoder and certain subcarriers can be nulled for the insertion of training pilots. These pilot positions are optimally selected to minimize the distortion of the transmitted time-domain signal (OFDM modulated signal) caused by the aforementioned subcarrier-removal at the corresponding pilot positions. The associated new receiver structure is also presented, where three different blind pilot-detection techniques are designed without any a priori knowledge of the pilot positions (based on sample variance, subspace decomposition, and Jarqur-Bera (JB) statistics, respectively), and the distorted data symbols can thus be iteratively reconstructed. Besides, rigorous theoretical analysis and Monte Carlo simulation results both demonstrate that our proposed new OFDM system using dynamical pilot positions is more robust than the conventional OFDM system using the fixed pilot positions over multipath fading channels.

The optimal data-fusion rule was first established for multiple-sensor detection systems in 1986. Most subsequent works have been focused on the corresponding implementation aspects. The probability of false alarm and the probability of miss detection used in this data-fusion rule are quite difficult to precisely enumerate in practice. Although the improved data-fusion implementation techniques are now available, most existing cooperative spectrum-sensing techniques are still based on the simple energy-detection algorithm, which is prone to failure in many scenarios. In this paper, we propose a novel adaptive cooperative spectrum-sensing scheme based on our recently proposed single-reception spectrum-sensing technique. We also found that the commonly-used sample-average estimator for the cumulative weights in the data-fusion rule becomes unreliable in time-varying environments. To overcome this drawback, we adopt a temporal discount factor, which is crucial to the probability estimators. New theoretical analysis to justify the advantage of our proposed new estimators over the conventional sample-average estimators and to determine the optimal numerical value of the proposed discount factor is presented. The Monte Carlo simulation results are also provided to demonstrate the superiority of our proposed adaptive cooperative spectrum-sensing method in both stationary and time-varying environments.

Adaptive modulation and coding (AMC) is commonly used in wireless systems to dynamically change the modulation and coding schemes (MCSs) in subsequent frames such that the spectral efficiency can be adapted to various channel conditions. The spectrum and energy efficiency would decrease if the adopted MCS option at the transmitter needs to be dynamically transmitted to the receiver through a secure control channel. To combat this problem, in this paper, we would like to propose a novel blind channel-encoder identification scheme with low computational complexity for Reed-Solomon (RS) codes over Galois field GF(q), which could also be applied to other similar non-binary channel codes as well. Our proposed new scheme involves the estimation of the channel parameters using the expectation-maximization (EM) algorithm, the calculation of the log-likelihood ratio vectors (LLRVs) of the syndrome a posteriori probabilities over GF(q), and the identification of the non-binary RS encoder in use subject to the maximum average log-likelihood ratio (LLR) over the pre-selected candidate encoder set. Simulation results justify the effectiveness of this new mechanism.

The optimal data-fusion rule was first established for multiple-sensor detection systems in 1986. The probability of false alarm and the probability of miss detection required in this data-fusion rule are quite difficult to precisely enumerate in practice. Although the improved data-fusion implementation techniques are available, most existing cooperative spectrum-sensing techniques are still based on the simple energy-detection algorithm, which is prone to failure in many scenarios. In our previous paper, we proposed a novel adaptive cooperative spectrum-sensing scheme based on Jarque-Bera (JB) statistics. However, the commonly-used sample-average estimator for the cumulative weights becomes unreliable in time-varying environments. To overcome this drawback, in this paper, we adopt a temporal discount factor, which is crucial to the probability estimators. New theoretical analysis to justify the advantage of our proposed new estimators over the conventional sample-average estimators and to determine the optimal numerical value of the proposed discount factor is presented. The Monte Carlo simulation results are also provided to demonstrate the superiority of our proposed adaptive cooperative spectrum sensing method in time-varying environments.

The finite impulse response (FIR) filter design has been a hot topic over many decades due to its guaranteed stability and wide variety of applications. The simplest way to design an FIR low-pass filter is to truncate the infinite-long sine sequence. This direct approach is a special case of window-FIR filter design with the rectangular window. Other windows were also proposed to deal with this filter design problem. Alternatively, given a fixed filter (window) length, optimization techniques can be employed to design the best window sequences. In this paper, we would like to present a novel computationally-efficient optimal window design algorithm for the existing window-FIR approach. Since the signal processing storage devices become less costly and more powerful, designing long FIR filters becomes prevalent in the modern telecommunication and signal processing applications. Therefore, the computationally-efficient filter design schemes are in urgent demand. Our proposed technique is based on a fast eigen-decomposition algorithm. The computational complexity of our method is O(N2log(N)) compared to O(N3) of the conventional method.

The optimal data fusion rule for multiple sensor detection systems based on the Bayesian criterion has been derived by Chair and Varshney in 1986. However, most of the following works are focused on how to implement such a fusion rule, since the probability of false alarm and the probability of miss detection are hard to evaluate in practice. Till now, although more and more satisfactory data-fusion implementation schemes are available, most of the cooperative spectrum sensing techniques are based on the simple energy-detection algorithm, which only relies on the energy of the received signal. However, when noise is relatively large or the time-varying characteristics of the signal are conspicuous, the energy-detection spectrum sensing algorithm is more prone to fail. Thus, in this paper, we propose a new adaptive cooperative spectrum sensing scheme, which is based on a novel detection algorithm involving JB (Jarque-Bera) statistic. The ROC (receiver-operating characteristic) curves show that our new cooperative spectrum sensing scheme is more robust than that based on the energy-detection spectrum sensing algorithm. Besides, the performance comparison also implies that the optimal data-fusion rule in our new cooperative spectrum sensing scheme is superior to the commonly adopted “OR” and “AND” rules in the existing literature.

The digital imaging technology has grown explosively for multimedia applications in recent years. The need for the copyrighted digitalized media becomes urgent nowadays. An approach for the digital copyright protection is to employ advanced watermarking techniques, where watermarks can reveal the ownership identities. Generally speaking, the watermarks are embedded into image or video signals. In this paper, we will investigate digital watermarking techniques and propose a new optimal watermarking scheme. When multiple embedded watermarks are considered, a new analysis for the signal-to-interference-plus-noise-ratios (SINRs) with respect to the subject signal and the watermark signals is carried out. The objective quality measure for the digital watermarking applications should essentially consist of both signal-to-interference-plus-noise-ratio for the subject signal and similarity coefficients for the watermarks. In order to optimize the aforementioned objective measure, we design a novel efficient scale-factor optimization scheme, which can lead to the maximum overall SINR for both subject signal and watermarks. Simulation results are also demonstrated to illustrate the effectiveness of our proposed method.