Most of the existing research focuses on the recognition of micro-expressions, and few studies how to recognize the action units of micro-expressions. This is due to the low intensity of the facial action unit, which is not easily to be recognized. To solve this problem, we proposed a micro-expression action unit recognition algorithm based on dynamic image and spatial pyramids. First, the video is passed through the dynamic image generation module to generate a dynamic image and extract the motion information contained in all frames. Then, given the subtle movement properties of micro-expressions, different levels of semantic features are obtained through spatial pyramids. It is also known that micro-expressions appear in the small range and are concentrated in local area of the face, so the regional feature network and attention mechanism are used for the image features of each layer. Finally, due to the weak correlation between each action unit, our models are trained separately. Experiments on CASME and CAS(ME)2 datasets verify that the proposed algorithm has shown better action unit recognition performance compared with other advanced methods.

Most of the existing research focuses on the recognition of microexpressions, and few studies how to recognize the action units of micro-expressions. This is due to the low intensity of the facial action unit, which is not easily to be recognized. To solve this problem, we proposed a micro-expression action unit recognition algorithm based on dynamic image and spatial pyramids. First, the video is passed through the dynamic image generation module to generate a dynamic image and extract the motion information contained in all frames. Then, given the subtle movement properties of micro-expressions, different levels of semantic features are obtained through spatial pyramids. It is also known that micro-expressions appear in the small range and are concentrated in local area of the face, so the regional feature network and attention mechanism are used for the image features of each layer. Finally, due to the weak correlation between each action unit, our models are trained separately. Experiments on CASME and CAS(ME) 2 datasets verify that the proposed algorithm has shown better action unit recognition performance compared with other advanced methods.

No abstract

Deceptive behaviour is a common phenomenon in human society. Research has shown that humans are not good at distinguishing deception, so studying automated deception detection techniques is a critical task. Most of the relevant technologies are susceptible to personal and environmental influences: EEG-based technologies need large and expensive equipment, facial-based technologies are sensitive with the camera’s perspective, and these reasons have somewhat limited the development of applications for deception detection technologies. In contrast, the equipment required for speech deception detection is cheap and easy to use, and the capture of speech is highly covert. Based on the application of signal decomposition algorithms in other fields such as EEG signals and speech emotion recognition, this paper proposed a signal decomposition and reconstruction method based on EMD to process the speech signal and a better deception detection performance was obtained by improving the speech quality. The comparison results with other decomposition algorithms showed that the EMD decomposition algorithm is the most suitable for our method. Across many different classification algorithms, accuracy improved by an average of 2.05% and the F1 score improved by an average of 1.7%. In addition, a new deception detector, called the TCN-LSTM network, was proposed in this paper. Experiments showed that this network organically combines the processing capability of TCN and LSTM for time series data; the recognition rate of deception detection was greatly improved, with the highest accuracy and F1 score reaching 86.2% and 86.0% under the EMD-based signal decomposition reconstruction method. Based on the research in this paper, the signal decomposition algorithms need to be further optimised for speech signals and more classification algorithms not used for this task should be tried.

A practical solution to the power allocation problem in ultra-dense small cell networks can be achieved by using deep reinforcement learning (DRL) methods. Unlike traditional algorithms, DRL methods are capable of achieving low latency and operating without the need for global real-time channel state information (CSI). Based on the actor–critic framework, we propose a policy optimization of the power allocation algorithm (POPA) for small cell networks in this paper. The POPA adopts the proximal policy optimization (PPO) algorithm to update the policy, which has been shown to have stable exploration and convergence effects in our simulations. Thanks to our proposed actor–critic architecture with distributed execution and centralized exploration training, the POPA can meet real-time requirements and has multi-dimensional scalability. Through simulations, we demonstrate that the POPA outperforms existing methods in terms of spectral efficiency. Our findings suggest that the POPA can be of practical value for power allocation in small cell networks.