A Projection-Free Adaptive Momentum Optimization Algorithm for Mobile Multimedia Computing

Mobile multimedia applications have become an integral part of modern digital life, offering services such as video streaming, online gaming, virtual reality, and real-time communication. Despite their growing popularity, these applications face inherent challenges due to the limited computational resources, storage capacity, and battery life of mobile devices. Cloud computing has emerged as a transformative solution, providing scalable processing power, extensive storage, and on-demand services. This paper presents a theoretical study on the performance evaluation of mobile multimedia applications in cloud computing environments. The study focuses on key performance metrics including latency, throughput, energy consumption, scalability, and quality of service (QoS). By critically analyzing the impact of cloud integration on mobile multimedia applications, this research identifies the benefits, limitations, and optimization strategies required to improve application performance. The findings highlight the significance of cloud-based architectures, task offloading, and adaptive resource management in achieving efficient and reliable mobile multimedia services.

The major preoccupation regarding the actual research is to visualize a theoretical model of the natural forces interaction between a multimedia mobile application deployed upon an educational activity over a group of students and an automatic process of evaluation regarding the final academic results. We implemented a new metric indicator, named here with effectiveness of a multimedia mobile application, correlated with the educational resultant force emerged from a mobile environment learning experiment. The variables involved in our model are connected to the cross-disciplinary curricula, the educational time coverage disclosed by a multimedia mobile application, the multimedia concentration, and the academic performance measured just in time at the interface between formal and informal evaluation. We are proposing another educational force couple, a resultant force, which is represented by the quality attributes allocated to the mobile devices and, furthermore, to the multimedia mobile applications. We are using the assumption that a multimedia mobile application generates an educational field in front of a learner taking into consideration not only the easiness of the process by which the learner’s attention is retaining toward a mobile device. The fact that between a mobile user and a multimedia mobile application could be established a quantitative relation of physical attraction or rejection, disclosed by surveys and questionnaires including qualities attributes, was an argument for our hypothesis that educational forces encountered could be explained by a cloaking educational field where a touch screen electric impulse is only the beginning of the educational magnetism encountered at the learners.KeywordsMobile applicationsMobile learningAndroidImpact on academic performance

Human activity recognition using sensor data has become a research hotspot in the field of ubiquitous computing and has a wide range of application scenarios in real life. Effectively extracting and fusing the rich spatiotemporal information contained in the sensor data from various sensors and multiple wearable devices is a challenging task. Moreover, for practical deployment applications, the limited computational and storage resources of edge devices also impose requirements on the model size and inference speed. Hence, we propose STFNet, an enhanced and lightweight spatiotemporal fusion network for wearable human activity recognition (WHAR), in which we innovatively divide the extraction and fusion process of spatiotemporal features into three stages, including sensing feature extraction (SFE), temporal feature extraction (TFE), and channel fusion (CF), to learn incrementally and reduce information loss. We first extract shallow intrachannel features and learn local interchannel information using a lightweight local attention mechanism. Then, the learned features are fed into a hierarchical convolution structure with residual connections to extract multiscale and comprehensive temporal features. Finally, we use graph convolutions to further fuse the deep interchannel information. By replacing the conventional global attention mechanism with a local one and using hierarchical convolution to replace dilated or multikernel convolution, STFNet achieves the purpose of lightweight design. In the experiments, STFNet performs better on public datasets than the selected WHAR models published in recent years and demonstrates its lightweight characteristics on edge device deployment.

In order to solve the problems of disrupting aviation order, stealing private information by the abuse of commercial small drones, even terrorist attacks caused by drones, many drone target detection algorithms based on deep learning have been proposed in recent years, but due to the lack of large-scale public datasets, most of the existing researches are based on self-made datasets, the algorithms lack comparative analysis. In this paper, we summarize and analyze the current research status of intrusion drone target detection based on deep learning. The paper summarizes the existing target detection algorithms based on deep learning, aiming at the problem that public drone target datasets are insufficient and difficult to obtain, it summarizes the currently publicly available drone datasets, then summarizes the intrusion drone target detection algorithms based on deep learning in recent years. Finally, the detection performance of some algorithms under different public datasets is compared, the difficulties in current research and direction of next research are analyzed.

High-data-rate mobile multimedia applications can greatly increase energy consumption, leading to an emerging trend of addressing the “energy efficiency” aspect of mobile networks. Cognitive mobile networks with small cells are important techniques for meeting the high-data-rate requirements and improving the energy efficiency of mobile multimedia communications. However, most existing works do not consider the power grid, which provides electricity to mobile networks. Currently, the power grid is experiencing a significant shift from the traditional grid to the smart grid. In the smart grid environment, only considering energy efficiency may not be sufficient since the dynamics of the smart grid will have significant impacts on mobile networks. In this paper, we study green cognitive mobile networks with small cells in the smart grid environment. Unlike most existing studies on cognitive networks, where only the radio spectrum is sensed, our cognitive networks sense not only the radio spectrum environment but also the smart grid environment, based on which power allocation and interference management for multimedia communications are performed. We formulate the problems of electricity price decision, energy-efficient power allocation, and interference management as a three-stage Stackelberg game. A homogeneous Bertrand game with asymmetric costs is used to model price decisions made by the electricity retailers. A backward induction method is used to analyze the proposed Stackelberg game. Simulation results show that our proposed scheme can significantly reduce operational expenditure and CO2 emissions in cognitive mobile networks with small cells for multimedia communications.

Demand for mobile video applications is growing today in wireless handheld platforms. Optimizing instruction set architectures and employing SIMD techniques is a logical approach towards attaining higher performance in mobile multimedia applications. Intel® Wireless MMX� technology has been designed to accelerate mobile multimedia and applications processing in a power efficient manner. This paper provides an overview of Intel® Wireless MMX� technology, a 64-bit Single Instruction Multiple Data (SIMD) coprocessor for the Intel® XScale® microarchitecture, and the key features of the architecture that specifically enhance the multi-media performance. Tools and techniques for optimization are also described.

During the last decade, new mobile multimedia applications have emerged for mobile and network multimedia, wireless multimedia communication, audio/video teleconferencing, remote assistance, digital storage systems, secure audio transmission and so on. In order to meet these requirements, tremendous research efforts have been put in the development of efficient digital audio coding technologies. In China, AVS-M audio standard is such an audio technology targeting for mobile multimedia applications which is developed and owned by China Audio and Video Coding Standard Workgroup. In this paper, AVS-M audio standard is discussed by revealing the technical principles of the en- and decoding, the standardization situation and the suitability of the codec in relation to technology available, economical feasibility and the market needs. Finally it concludes with a brief discussion of future research directions.

Computational resources in a mobile terminal and in wireless transmission channels set strict requirements for audio coding in mobile multimedia applications. This paper introduces, a low complex audio encoding method enabling high-quality audio for mobile multimedia. The presented method is the low complexity version of the extended AMR-WB audio encoder in the 3GPP standards. The encoder was designed based on requirements of the 3GPP multimedia messaging service (MMS)

Due to improper selection of gradient update direction or learning rate, SGD optimization algorithms for deep learning suffer from oscillation and slow convergence. Although Adam algorithm can adaptively adjust the update direction and learning rate at the same time, it still has the overshoot phenomenon, and hence suffers from wasting computing resources and slow convergence. In this work, the PID controller from the feedback control area is borrowed to re‐express the adaptive optimization algorithm (the Adam optimization algorithm is derived into the integral I component form) of deep learning. In order to alleviate the overshoot phenomenon and hence speed up the convergence of Adam, a complete adaptive PID optimizer (adaptive‐PID) is proposed by incorporating the proportional P and derivative D component. Extensive experiments on standard data sets verify that the proposed adaptive‐PID algorithm significantly outperforms Adam algorithm in terms of convergence rate and accuracy.

Forest fires pose a serious threat to the environment with the potential of causing ecological harm, financial losses, and human casualties. While research suggests that climate change will increase the frequency and severity of these fires, recent developments in deep learning and convolutional neural networks (CNN) have greatly enhanced fire detection techniques and capability. These models can be leveraged by unmanned aerial vehicles (UAVs) to automatically monitor burning areas. However, drones can carry only limited computational and power resources; therefore, on-board computing capabilities are constrained by hardware limitations. This work focuses on the design of segmentation models to identify and localize active burning areas from aerial RGB images processed on limited computing resources. To achieve this goal, the research compares the performance of different variants of the DeepLabv3 neural network model for fire segmentation when trained and tested with the FLAME dataset using a k-fold cross validation approach. Experimental results are compared with U-Net, a benchmark model used with the FLAME dataset, by implementing this model in the same codebase as the DeepLabv3 model. This work demonstrates that a refined version of DeepLabv3, with a MobileNetv2 backbone using pretrained layers and a simplified atrous spatial pyramid pooling (ASPP) module, yields a similar performance to U-Net, with a precision of 87.8% and a recall of 83.2%, while only requiring 20% of the number of parameters involved with the U-Net topology. This significantly reduces memory and power consumption, enabling longer UAV flight duration and reducing the processing overhead associated with sensor input, making it more suitable for deployment on unmanned aerial vehicles. The model’s compact architecture, implemented using TensorFlow and Keras for model design and training, along with OpenCV for image preprocessing, makes it portable and easy to integrate with edge devices such as NVIDIA Jetson boards.

With the continuous progress of UAV technology and the rapid development of mobile edge computing (MEC), the UAV-assisted MEC system has shown great application potential in special fields such as disaster rescue and emergency response. However, traditional deep reinforcement learning (DRL) decision-making methods suffer from limitations such as difficulty in balancing multiple objectives and training convergence when making mixed action space decisions for UAV path planning and task offloading. This article innovatively proposes a hybrid decision framework based on the improved Dynamic Adaptive Genetic Optimization Algorithm (DAGOA) and soft actor–critic with hierarchical action decomposition, an uncertainty-quantified critic ensemble, and adaptive entropy temperature, where DAGOA performs an effective search and optimization in discrete action space, while SAC can perform fine control and adjustment in continuous action space. By combining the above algorithms, the joint optimization of drone path planning and task offloading can be achieved, improving the overall performance of the system. The experimental results show that the framework offers significant advantages in improving system performance, reducing energy consumption, and enhancing task completion efficiency. When the system adopts a hybrid decision framework, the reward score increases by a maximum of 153.53% compared to pure deep reinforcement learning algorithms for decision-making. Moreover, it can achieve an average improvement of 61.09% on the basis of various reinforcement learning algorithms such as proposed SAC, proximal policy optimization (PPO), deep deterministic policy gradient (DDPG), and twin delayed deep deterministic policy gradient (TD3).

Emerging as one of the most contemporary machine learning techniques, deep learning has shown success in areas such as image classification, speech recognition, and even playing games through the use of hierarchical architecture which includes many layers of non-linear information. In this paper, a powerful deep learning pipeline, intelligent deep learning (iDeepLe) is proposed for both regression and classification tasks. iDeepLe is written in Python with the help of various API libraries such as Keras, TensorFlow, and Scikit-Learn. The core idea of the pipeline is inspired by the sequential modeling with considering numerous layers of neurons to build the deep architecture. Each layer in the sequential deep model can perform independently as a module with minimum finitudes and does not limit the performance of the other layers. iDeepLe has the ability of employing grid search, random search, and Bayesian optimization to tune the most significant predictor input variables and hyper-parameters in the deep model via adaptive learning rate optimization algorithms for both accuracy and complexity, while simultaneously solving the unknown parameters of the regression or the classification model. The parallel pipeline of iDeepLe has the capacity to handle big data problems using Apache Spark, Apache Arrow, High Performance Computing (HPC) and GPU-enabled machines as well. In this paper, to show the importance of the optimization in deep learning, an exhaustive study of the impact of hyper-parameters in a simple and a deep model using optimization algorithms with adaptive learning rate was carried out.

Breast cancer is the most prevalent cancer and the second cause of cancer related death among women in the United States. Accurate and early detection of breast cancer can reduce the number of mortalities. Recent works explore deep learning techniques with ultrasound for detecting malignant breast lesions. However, the lack of explanatory features, need for segmentation, and high computational complexity limit their applicability in this detection. Therefore, we propose a novel ultrasound-based breast lesion classification framework that utilizes two-dimensional variational mode decomposition (2D-VMD) which provides self-explanatory features for guiding a convolutional neural network (CNN) with mixed pooling and attention mechanisms. The visual inspection of these features demonstrates their explainability in terms of discriminative lesion-specific boundary and texture in the decomposed modes of benign and malignant images, which further guide the deep learning network for enhanced classification. The proposed framework can classify the lesions with accuracies of 98% and 93% in two public breast ultrasound datasets and 89% in an in-house dataset without having to segment the lesions unlike existing techniques, along with an optimal trade-off between the sensitivity and specificity. 2D-VMD improves the areas under the receiver operating characteristics and precision-recall curves by 5% and 10% respectively. The proposed method achieves relative improvement of 14.47%(8.42%) (mean (SD)) in accuracy over state-of-the-art methods for one public dataset, and 5.75%(4.52%) for another public dataset with comparable performance to two existing methods. Further, it is computationally efficient with a reduction of in floating point operations as compared to existing methods.

COVID-19, an infectious coronavirus disease, caused a pandemic with countless deaths. From the outset, clinical institutes have explored computed tomography as an effective and complementary screening tool alongside the reverse transcriptase-polymerase chain reaction. Deep learning techniques have shown promising results in similar medical tasks and, hence, may provide solutions to COVID-19 based on medical images of patients. We aim to contribute to the research in this field by: (i) Comparing different architectures on a public and extended reference dataset to find the most suitable; (ii) Proposing a patient-oriented investigation of the best performing networks; and (iii) Evaluating their robustness in a real-world scenario, represented by cross-dataset experiments. We exploited ten well-known convolutional neural networks on two public datasets. The results show that, on the reference dataset, the most suitable architecture is VGG19, which (i) Achieved 98.87% accuracy in the network comparison; (ii) Obtained 95.91% accuracy on the patient status classification, even though it misclassifies some patients that other networks classify correctly; and (iii) The cross-dataset experiments exhibit the limitations of deep learning approaches in a real-world scenario with 70.15% accuracy, which need further investigation to improve the robustness. Thus, VGG19 architecture showed promising performance in the classification of COVID-19 cases. Nonetheless, this architecture enables extensive improvements based on its modification, or even with preprocessing step in addition to it. Finally, the cross-dataset experiments exposed the critical weakness of classifying images from heterogeneous data sources, compatible with a real-world scenario.

In public areas, crowd stampedes and incidents generate huge negative impacts on public security. Accurate and efficient crowd density estimation is critical to monitor crowd status for developing evacuation strategies. The existing crowd density estimation methods are established based on complex deep-learning algorithms which are usually more accurate, but, on the other side, they require much more computational resources. Consequently, cloud-computing is the only option for deploying crowd density estimation algorithms, which needs tremendous resources for real-time video data transmission and the malfunction and delay of internet service may cause wrong and delayed estimation results. Edge computing is a novel concept of accomplishing computing tasks only relying on the computational resources of edge devices. Estimating crowd density on edge devices rather than conveying images to cloud server for further analysis has several advantages, including 1) reducing network bandwidth pressure from remote transmission; 2) avoiding risks of leaking privacy on images; and 3) improving computation efficient. This study designs an edge computing-enabled crowd density estimation model based on the residual bottleneck block and dilated convolution. The experiments are designed and conducted on public crowd data sets to verify accuracy, storage cost and computation efficiency of our model. According to the experimental results, the proposed model achieves a considerable improvement in operational efficiency, while keep the accuracy at the same level with the complex deep-learning algorithms. Furthermore, the proposed model is implemented on a real edge device to detect real-world crowd density in a Beijing subway station.