Search

Abstract Understanding human behavior and human action recognition are both essential components of effective surveillance video analysis for the purpose of guaranteeing public safety. However, existing approaches such as three-dimensional convolutional neural networks (3D CNN) and two-stream neural networks (2SNN) have computational hurdles due to the significant parameterization they require. In this paper, we offer HARNet, a specialized lightweight residual 3D CNN that is built on directed acyclic graphs and was created expressly to handle these issues and achieve effective human action detection. The suggested method presents an innovative pipeline for creating spatial motion data from raw video inputs, which makes successful latent representation learning of human motions easier to accomplish. This generated input is then supplied into HARNet, which processes spatial and motion information in a single stream in an effective manner, maximizing the benefits of both types of cues. The use of traditional machine learning classifiers is done in order to further improve the discriminative capacity of the features that have been learned. To be more specific, we use the latent representations that are stored in HARNet’s fully connected layer and use them as our deep learnt features. After that, these features are entered into the Support Vector Machine (SVM) classifier in order to accomplish action recognition. In order to evaluate the HARNet-SVM method that was developed, empirical tests were run on commonly used action recognition datasets such as UCF101, HMDB51, and the KTH dataset. These tests were carried out in order to gather data for the evaluation. The experimental results show that our method is superior to other state-of-the-art approaches, achieving considerable performance increases of 2.75% on UCF101, 10.94% on HMDB51, and 0.18% on the KTH dataset. These results were obtained by running the method on each dataset separately. Our findings demonstrate the usefulness of HARNet’s lightweight design and highlight the significance of utilizing SVM classifiers with deep learnt features for the purpose of accurate and computationally efficient human activity recognition in surveillance videos. This work helps to the advancement of surveillance technology, which in turn makes video analysis in applications that take place in the real world safer and more dependable.

Filter Bank Multicarrier (FBMC) is considered one of the strong applicants for a radio system beyond the fifth generation (B5G) that improves spectral access and lowers interference. It utilizes a prototype filter for each sub-carrier, making it best for the beyond fifth generation (B5G) framework. The performance of the FBMC is hugely impacted by the high peak-to-average power ratio (PAPR), which lowers the effectiveness of the power amplifier (PA) used in the 5G-based FBMC waveform. The conventional partial transmission sequence (PTS) technique requires high computational complexity due to the need for multiple Inverse Fast Fourier Transforms (IFFTs) and phase optimization, which can increase processing time and system latency. This article proposes a hybrid method combining a partial transmission sequence and recurrent neural network (RNN) known as PTS-RNNs. RNNs improve the performance of the PTS by efficiently predicting optimal phase factors, reducing computational complexity, and lowering the PAPR of the FBMC waveform. The parameters such as PAPR, bit error rate (BER), and power spectral density (PSD) are estimated for 256 sub-carriers under the Rayleigh and Rician channels for FBMC and orthogonal frequency division multiplexing (OFDM). The experiment results reveal that the proposed PTS-RNNs method achieves an efficient 55.45 % and 67.56 % power saving performance for Rayleigh and Rician channels, with enhanced PSD performance while preserving the BER compared to the traditional selective mapping (SLM) and PTS methods. It is also noticeable that by adding more sub-blocks and phase parameters, PAPR can be further optimised.

Abstract The diversity of consumer demand for take-out food has led to the characteristic structure of one order with multiple items, where the different food items in a single order are provided by two or more merchants. In the context of multi-item delivery for take-out orders with time windows, this study investigates vehicle routing for order delivery. This research aims to improve the service level of merchants and the efficiency of delivery vehicles. Food vendors receive orders from consumers through online platforms, then package the food items according to the orders. This method is a preliminary exploration of the issue of fulfilment of takeout orders on online platforms, and can provide preliminary theoretical support for decision-making on the delivery process of takeout orders on online platforms. In the context of online catering sales platforms and offline food sales merchants, this paper studies the delivery problem of takeaway orders with a time window and the characteristics of one order and multiple items, and constructs a method that takes the order delivery time requirements into account and minimizes the total order fulfilment cost. Taking into account the time window constraints of both physical restaurants and consumers, genetic algorithms are utilized to solve the order delivery problem. Finally, through case studies and experiments, the effectiveness and feasibility of the mathematical model are validated. Practical recommendations and insights are provided from the perspective of management and route planning.

Abstract The current study focuses on examining the characteristics of biofuel obtained from the pyrolysis of Madhuca longifolia residues, since the selected forest residue was primarily motivated by its greater volatile matter content. The study used several analytical techniques to describe pyrolysis oil, char, and gas obtained from slow pyrolysis process conducted between 350 and 600 °C in a fixed-bed reactor. Initially, the effect of process temperature on product distribution was assessed to motivate maximum pyrolysis oil yield and found to be 44.2 wt% at pyrolysis temperature of 475 °C, while the yields of char and gas were 22.1 wt% and 33.7 wt%, respectively. In order to determine the suitability of the feedstock, the Madhuca longifolia residues were analyzed by TGA and FT-IR, which revealed that the feedstock could be a feasible option as an energy source. The characterization of pyrolysis oil, char, and gas has been done through various analytical methods like FT-IR, GC-MS, and gas chromatography. The physicochemical characteristics of the pyrolysis oil sample were examined, and the results showed that the oil is a viscous liquid with a lower heating value than conventional diesel. The FT-IR and GC-MS analysis of pyrolysis oil revealed the presence of increased levels of oxygenated chemicals, acids, and phenol derivatives. The findings of the FT-IR analysis of char indicated the existence of aromatic and aliphatic hydrocarbons. The increased carbon content in the char indicated the possibility of using solid fuel. Gas chromatography was used to examine the chemical structure of the pyrolysis gas, and the results showed the existence of combustible elements.

Diagnosing brain tumors is difficult, especially at an early stage of the disease. Conventional approaches often cause delays in providing required treatment to the patients and shorten their lifespan. This paper presents a novel integrated approach with advanced subsets of artificial intelligence, including deep learning and supervised learning algorithms. These new technologies have demonstrated outstanding potential due to their ability to capture the appropriate features based on the input data. They can assist medical experts in identifying abnormal growth of cells inside the brain. We use publicly available brain magnetic resonance imaging (MRI) datasets to diagnose brain tumors and develop an automated system. The proposed approach uses data augmentation to enhance the image sizes and maintain standardization. We then deploy a visual geometry group with 16 layers following transfer learning to help minimize the medical experts’ workload in making accurate decisions. We extract the most significant features and improve the diagnostic speed and accuracy using a supervised learning algorithm and linear support vector machines (SVM). The proposed model outperforms the existing approaches with an accuracy of 98.87%, precision of 99.09%, recall of 98.73%, specificity of 99.02%, and F1-score of 98.91%.

The integration of sensor technology and automated analytics in wearable devices marks a significant advancement in sports science, aiming to proactively prevent injuries and optimize athletic performance. This paper delves into the application of advanced sensor systems and automated data analysis tools in the realm of athletics. These wearables, equipped with a multitude of sensors, continuously monitor an athlete's physiological and biomechanical data, including heart rate, muscle activity, and movement dynamics. The automated processing of this data through machine learning algorithms enables real-time insights and predictive analytics, offering a novel approach to injury prevention. By analyzing the intricate patterns of an athlete's performance metrics, these sensor-augmented wearables facilitate a data-driven strategy for reducing risks of soft-tissue and heat-related injuries. This technology empowers coaches and athletes with actionable insights, fostering a safer and more efficient training environment. Our exploration showcases the transformative impact of sensor-enhanced wearables and automated analytics in elevating the standards of injury prevention and performance optimization in sports.

Heart strokes are a significant global health concern, profoundly affecting the wellbeing of the population. Many research endeavors have focused on developing predictive models for heart strokes using ML and DL techniques. Nevertheless, prior studies have often failed to bridge the gap between complex ML models and their interpretability in clinical contexts, leaving healthcare professionals hesitant to embrace them for critical decision-making. This research introduces a meticulously designed, effective, and easily interpretable approach for heart stroke prediction, empowered by explainable AI techniques. Our contributions include a meticulously designed model, incorporating pivotal techniques such as resampling, data leakage prevention, feature selection, and emphasizing the model’s comprehensibility for healthcare practitioners. This multifaceted approach holds the potential to significantly impact the field of healthcare by offering a reliable and understandable tool for heart stroke prediction. In our research, we harnessed the potential of the Stroke Prediction Dataset, a valuable resource containing 11 distinct attributes. Applying these techniques, including model interpretability measures such as permutation importance and explainability methods like LIME, has achieved impressive results. While permutation importance provides insights into feature importance globally, LIME complements this by offering local and instance-specific explanations. Together, they contribute to a comprehensive understanding of the Artificial Neural Network (ANN) model. The combination of these techniques not only aids in understanding the features that drive overall model performance but also helps in interpreting and validating individual predictions. The ANN model has achieved an outstanding accuracy rate of 95%.