The goal of this paper is to introduce robust signalprocessing techniques for realistic communication scenarios with fading in ultra-dense massive MIMO communications. The proposed techniques tackle major challenges ...
详细信息
Graph Meta-learning methods have improved the performance of few-shot node classification by means of applying meta-learning to the data in non-Euclidean domains. However, most works focus on adopting a single domain,...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Graph Meta-learning methods have improved the performance of few-shot node classification by means of applying meta-learning to the data in non-Euclidean domains. However, most works focus on adopting a single domain, ignoring the fact that tasks in various domains may be distinct, which can cause overfitting problems and thus limit generalizability. To tackle this challenge, we propose a novel Graph Meta-learning framework called Feature-Enhanced Cross-domain Graph Meta-learning that consists of two crucial modules: 1) A feature information extraction module that aims to capture discriminative node importance and simulate various node feature distributions under distinct domains; 2) A heterogeneous graph encoder module that leverages the enhanced node features and topological information to generate task-specific node embeddings with simple fine-tuning. Moreover, we meta-learn the parameters involved to ensure the generalizability in the unseen domains. Results show that our method markedly outperforms the existing state-of-the-art methods.
This paper explores the advancements in human motion-tracking technology within the past three years, analyzing research that presents novel methods for capturing and analyzing human movement. We examine the methodolo...
详细信息
ISBN:
(数字)9798350362572
ISBN:
(纸本)9798350362589
This paper explores the advancements in human motion-tracking technology within the past three years, analyzing research that presents novel methods for capturing and analyzing human movement. We examine the methodologies employed by each approach, including sensor-based systems, radar technology, Wi-Fi signal utilization, lightweight networks for real-time action recognition, and a system for joint camera localization and human motion tracking. This paper compares each surveyed research's advantages, disadvantages, and limitations, highlighting their practical implications and potential applications. Finally, this review discusses the future directions and challenges within human motion tracking.
Fetal Phonocardiogram (fPCG) is a bio-signal obtained from recording the heart sounds from stethoscopes or auditory transducers for auscultations of FHR. Analysis of this data helps to give early diagnosis of any irre...
详细信息
ISBN:
(数字)9798350379525
ISBN:
(纸本)9798350379532
Fetal Phonocardiogram (fPCG) is a bio-signal obtained from recording the heart sounds from stethoscopes or auditory transducers for auscultations of FHR. Analysis of this data helps to give early diagnosis of any irregularity throughout Fetal development process. However, the analysis of this signal poses many challenges due to noise components like Motion Artifacts and Respiration of the Foetus and Mother, Uterine Contractions, Power Line Interference, Sensor Noise as well as Background Noise. Many approaches have been presented by the researchers to extract characteristic heart sound peaks from these signals to estimate the heart rate and to detect the abnormalities like arrhythmia. In this work, we present the results of application of advancedsignalprocessing techniques such as wavelet transform with adaptive thresholding, to extract the peaks of the fPCG signal. Simulations have been performed on the set of real data from fetus as well as simulated signals available in the resources. Simulation results show that Discrete Wavelet Transform with adaptive thresholding is a better technique for denoising the signal and isolating the heartbeat peaks as compared to other studies discussed in various resources. This study also includes evaluation of various parameters employed in the processes including the verification of peaks by setting up a window of specific time to analyze the peak count within each window for consistent peak extraction results. Various combinations of wavelets with different thresholding techniques have been studied and evaluated for accuracy.
Image processing is a very vital part of many medical diagnosis. With the advent of more technologically advanced devices, machine learning implementation has also proven to be boon in the medical world for imaging re...
详细信息
Image processing is a very vital part of many medical diagnosis. With the advent of more technologically advanced devices, machine learning implementation has also proven to be boon in the medical world for imaging related diagnosis. This paper aims to utilize different machine learning models to highlight its efficacy in the field of medical image analysis. The paper uses machine learning classifiers to classify breast cancer into malignant and benign using 31 attributes of the Wisconsin Breast Cancer Diagnostic dataset. Five different classifier models – Decision Trees, SVM, Naive Bayes, KNN and ANN were used to classify the tumors and it was observed and concluded that SVM model performed better with an accuracy of approximately 97.4% followed by ANN model with an accuracy of approximately 96.5%.
This paper presents the application of Artificial Intelligence (AI) techniques for quality inspection of transmission lines. Transmission lines play a crucial role in the reliable and efficient transfer of electrical ...
详细信息
ISBN:
(数字)9798350361186
ISBN:
(纸本)9798350361193
This paper presents the application of Artificial Intelligence (AI) techniques for quality inspection of transmission lines. Transmission lines play a crucial role in the reliable and efficient transfer of electrical power over long distances. However, faults and defects can occur due to various environmental and operational factors, leading to power outages and safety hazards. The proposed approach utilizes advanced Edge Deployable AI algorithms, including computer vision and Deep Learning, to automatically detect and classify potential faults and defects in transmission lines. The cutting edge high accuracy model is trained using i9 core, Nvidia RTX 3060 GPU with 6 GB vRAM, 16 GB RAM where we employ SSD - PyTorch based architecture and frameworks. The trained model is optimized to best accuracy and deployed over Nvidia Jetson Nano B01 using TensoRT Engine. A comprehensive dataset of images and videos of transmission lines is collected and labeled for training and testing the AI models. These models are trained to recognize patterns and anomalies indicative of various types of faults, such as broken conductors, insulator wear, and corrosion. The inspection process involves capturing images and videos of the transmission lines using drones or other surveillance systems. The captured data is then analyzed using the pre-trained AI models to identify potential issues. The AI algorithms can accurately detect and classify faults with high precision and recall, significantly reducing the need for manual inspection and minimizing human error. To enhance inspection accuracy, the AI models can be continuously updated and refined through a feedback loop using active learning algorithm.
The identification of traffic signs is a major challenge for intelligent automobiles. Recognition of traffic signs gives useful data, such as alerts and directions, for cooperative intelligent transport systems (CITS)...
详细信息
The use of the probability density distribution of random processes leads to certain difficulties in the implementation of signalprocessingalgorithms, especially for processing non-Gaussian processes. One of the adv...
详细信息
ISBN:
(纸本)9798350332636
The use of the probability density distribution of random processes leads to certain difficulties in the implementation of signalprocessingalgorithms, especially for processing non-Gaussian processes. One of the advanced approaches that allows describe non-Gaussian random processes is to use the moment and cumulant description of random variables. Based on this approach, two new methods of joint signal discrimination and parameter estimation are proposed. Nonlinear signalprocessing and taking into account the parameters of non-Gaussian processes can significantly improve the quality of signalprocessing compared to known classical methods.
In the realm of next-generation vehicle safety, this project introduces a cutting-edge approach to auto-stop functionality in modern vehicles. Leveraging the state-of the-art YOLO (You Only Look Once) algorithm, the s...
详细信息
ISBN:
(数字)9798350353068
ISBN:
(纸本)9798350353075
In the realm of next-generation vehicle safety, this project introduces a cutting-edge approach to auto-stop functionality in modern vehicles. Leveraging the state-of the-art YOLO (You Only Look Once) algorithm, the system focuses on enhancing driver safety through two critical aspects: facial identification and drowsiness monitoring. The YOLO algorithm efficiently detects and analyses facial features, providing real-time identification of the driver. Simultaneously, the system employs advanced drowsiness monitoring techniques, utilizing facial cues to gauge the driver's alertness levels. This futuristic safety initiative aims to mitigate potential accidents caused by driver fatigue or distraction by implementing an automatic auto-stop mechanism. The YOLO algorithm's speed and accuracy play a pivotal role in enabling swift and reliable detection of facial attributes, ensuring a seamless integration with the vehicle's safety system. By seamlessly integrating facial identification and drowsiness monitoring, the project presents a comprehensive solution to enhance road safety. The innovation lies in the project's ability to interpret facial expressions, track eye movements, and discern signs of drowsiness, enabling the vehicle to proactively intervene when a compromised driver state is detected. With an emphasis on real-time responsiveness, the system strives to revolutionize auto-stop technology, creating a safer driving environment for both the driver and other road users. This abstract encapsulates a groundbreaking fusion of advanced computer vision, machine learning, and automotive safety, ushering in a new era of proactive accident prevention in next-gen vehicles.
In today's globalized world, English communication skills are essential for career advancement and cross-cultural collaboration, enhancing access to information and opportunities. Automatic speech recognition, or ...
详细信息
ISBN:
(数字)9798350350654
ISBN:
(纸本)9798350350661
In today's globalized world, English communication skills are essential for career advancement and cross-cultural collaboration, enhancing access to information and opportunities. Automatic speech recognition, or ASR, is a separate machine-driven method for transcription and decoding spoken language. An ASR system typically uses a microphone to capture a speaker's audio input, analyze it using a model, algorithm, or pattern, and output the results, which are often text messages (Lai, Karat, Yankelovich, 2008). This paper provides a thorough method for utilizing Python-based tools and modules to extract and analyze linguistic characteristics from audio data. The suggested approach turns spoken language into text using voice recognition technology, and then it uses natural language processing (NLP) methods to extract different linguisticmetrics. Word count, sentence count, vocabulary size, average sentence length, average word length, sentiment score, speech pace, frequency of pauses, and average length of pauses are some of these measures. The technique also determines the speaker's speaking style. Keywords: audio analysis, linguistic features, natural language processing, pause detection, speech recognition, sentiment analysis, visualization.
暂无评论