Search results for: LIRA neural classifier

Authors: Ana Paula Soares, Alexandrina Lages, Helena Oliveira, Francisco-Javier Gutiérrez-Domínguez, Marisa Lousada

Abstract:

From an early age, exposure to a spoken language allows us to implicitly capture the structure underlying the succession of the speech sounds in that language and to segment it into meaningful units (words). Statistical learning (SL), i.e., the ability to pick up patterns in the sensory environment even without intention or consciousness of doing it, is thus assumed to play a central role in the acquisition of the rule-governed aspects of language and possibly to lie behind the language difficulties exhibited by children with development language disorder (DLD). The research conducted so far has, however, led to inconsistent results, which might stem from the behavioral tasks used to test SL. In a classic SL experiment, participants are first exposed to a continuous stream (e.g., syllables) in which, unbeknownst to the participants, stimuli are grouped into triplets that always appear together in the stream (e.g., ‘tokibu’, ‘tipolu’), with no pauses between each other (e.g., ‘tokibutipolugopilatokibu’) and without any information regarding the task or the stimuli. Following exposure, SL is assessed by asking participants to discriminate between triplets previously presented (‘tokibu’) from new sequences never presented together during exposure (‘kipopi’), i.e., to perform a two-alternative-forced-choice (2-AFC) task. Despite the widespread use of the 2-AFC to test SL, it has come under increasing criticism as it is an offline post-learning task that only assesses the result of the learning that had occurred during the previous exposure phase and that might be affected by other factors beyond the computation of regularities embedded in the input, typically the likelihood two syllables occurring together, a statistic known as transitional probability (TP). One solution to overcome these limitations is to assess SL as exposure to the stream unfolds using online techniques such as event-related potentials (ERP) that is highly sensitive to the time-course of the learning in the brain. Here we collected ERPs to examine the neurofunctional correlates of SL in preschool children with DLD, and chronological-age typical language development (TLD) controls who were exposed to an auditory stream in which eight three-syllable nonsense words, four of which presenting high-TPs and the other four low-TPs, to further analyze whether the ability of DLD and TLD children to extract-word-like units from the steam was modulated by words’ predictability. Moreover, to ascertain if the previous knowledge of the to-be-learned-regularities affected the neural responses to high- and low-TP words, children performed the auditory SL task, firstly, under implicit, and, subsequently, under explicit conditions. Although behavioral evidence of SL was not obtained in either group, the neural responses elicited during the exposure phases of the SL tasks differentiated children with DLD from children with TLD. Specifically, the results indicated that only children from the TDL group showed neural evidence of SL, particularly in the SL task performed under explicit conditions, firstly, for the low-TP, and, subsequently, for the high-TP ‘words’. Taken together, these findings support the view that children with DLD showed deficits in the extraction of the regularities embedded in the auditory input which might underlie the language difficulties.

Keywords: development language disorder, statistical learning, transitional probabilities, word segmentation

Procedia PDF Downloads 187

Authors: M. Sirhan, S. Bekhor, A. Sidess

Abstract:

In-service pavements deteriorate with time due to traffic wheel loads, environment, and climate conditions. Pavement deterioration leads to a reduction in their serviceability and structural behavior. Consequently, proper maintenance and rehabilitation (M&R) are necessary actions to keep the in-service pavement network at the desired level of serviceability. Due to resource and financial constraints, the pavement management system (PMS) prioritizes roads most in need of maintenance and rehabilitation action. It recommends a suitable action for each pavement based on the performance and surface condition of each road in the network. The pavement performance and condition are usually quantified and evaluated by different types of roughness-based and stress-based indices. Examples of such indices are Pavement Serviceability Index (PSI), Pavement Serviceability Ratio (PSR), Mean Panel Rating (MPR), Pavement Condition Rating (PCR), Ride Number (RN), Profile Index (PI), International Roughness Index (IRI), and Pavement Condition Index (PCI). PCI is commonly used in PMS as an indicator of the extent of the distresses on the pavement surface. PCI values range between 0 and 100; where 0 and 100 represent a highly deteriorated pavement and a newly constructed pavement, respectively. The PCI value is a function of distress type, severity, and density (measured as a percentage of the total pavement area). PCI is usually calculated iteratively using the 'Paver' program developed by the US Army Corps. The use of soft computing techniques, especially Artificial Neural Network (ANN), has become increasingly popular in the modeling of engineering problems. ANN techniques have successfully modeled the performance of the in-service pavements, due to its efficiency in predicting and solving non-linear relationships and dealing with an uncertain large amount of data. Typical regression models, which require a pre-defined relationship, can be replaced by ANN, which was found to be an appropriate tool for predicting the different pavement performance indices versus different factors as well. Subsequently, the objective of the presented study is to develop and train an ANN model that predicts the PCI values. The model’s input consists of percentage areas of 11 different damage types; alligator cracking, swelling, rutting, block cracking, longitudinal/transverse cracking, edge cracking, shoving, raveling, potholes, patching, and lane drop off, at three severity levels (low, medium, high) for each. The developed model was trained using 536,000 samples and tested on 134,000 samples. The samples were collected and prepared by The National Transport Infrastructure Company. The predicted results yielded satisfactory compliance with field measurements. The proposed model predicted PCI values with relatively low standard deviations, suggesting that it could be incorporated into the PMS for PCI determination. It is worth mentioning that the most influencing variables for PCI prediction are damages related to alligator cracking, swelling, rutting, and potholes.

Keywords: artificial neural networks, computer programming, pavement condition index, pavement management, performance prediction

Procedia PDF Downloads 137

Authors: Angelo Lerro, Manuela Battipede, Piero Gili, Alberto Brandl

Abstract:

Redundancy requirements for UAV (Unmanned Aerial Vehicle) are hardly faced due to the generally restricted amount of available space and allowable weight for the aircraft systems, limiting their exploitation. Essential equipment as the Air Data, Attitude and Heading Reference Systems (ADAHRS) require several external probes to measure significant data as the Angle of Attack or the Sideslip Angle. Previous research focused on the analysis of a patented technology named Smart-ADAHRS (Smart Air Data, Attitude and Heading Reference System) as an alternative method to obtain reliable and accurate estimates of the aerodynamic angles. This solution is based on an innovative sensor fusion algorithm implementing soft computing techniques and it allows to obtain a simplified inertial and air data system reducing external devices. In fact, only one external source of dynamic and static pressures is needed. This paper focuses on the benefits which would be gained by the implementation of this system in UAV applications. A simplification of the entire ADAHRS architecture will bring to reduce the overall cost together with improved safety performance. Smart-ADAHRS has currently reached Technology Readiness Level (TRL) 6. Real flight tests took place on ultralight aircraft equipped with a suitable Flight Test Instrumentation (FTI). The output of the algorithm using the flight test measurements demonstrates the capability for this fusion algorithm to embed in a single device multiple physical and virtual sensors. Any source of dynamic and static pressure can be integrated with this system gaining a significant improvement in terms of versatility.

Keywords: aerodynamic angles, air data system, flight test, neural network, unmanned aerial vehicle, virtual sensor

Procedia PDF Downloads 219

Authors: Rishabh Nagpal, Riya Uchagaonkar, Venkata Naga Narasimha Ashish Mernedi, Ahmed Hambaba

Abstract:

The rapid evolution of technology in the communication sector continually seeks to bridge the gap between different communities, notably between the deaf community and the hearing world. This project develops a comprehensive American Sign Language (ASL) recognition system, leveraging the advanced capabilities of convolutional neural networks (CNNs) and vision transformers (ViTs) to interpret and translate ASL in real-time. The primary objective of this system is to provide an effective communication tool that enables seamless interaction through accurate sign language interpretation. The architecture of the proposed system integrates dual networks -VGG16 for precise spatial feature extraction and vision transformers for contextual understanding of the sign language gestures. The system processes live input, extracting critical features through these sophisticated neural network models, and combines them to enhance gesture recognition accuracy. This integration facilitates a robust understanding of ASL by capturing detailed nuances and broader gesture dynamics. The system is evaluated through a series of tests that measure its efficiency and accuracy in real-world scenarios. Results indicate a high level of precision in recognizing diverse ASL signs, substantiating the potential of this technology in practical applications. Challenges such as enhancing the system’s ability to operate in varied environmental conditions and further expanding the dataset for training were identified and discussed. Future work will refine the model’s adaptability and incorporate haptic feedback to enhance the interactivity and richness of the user experience. This project demonstrates the feasibility of an advanced ASL recognition system and lays the groundwork for future innovations in assistive communication technologies.

Keywords: sign language, computer vision, vision transformer, VGG16, CNN

Procedia PDF Downloads 42

Authors: Colin Schmidt, Adrien Lecossier, Pascal Crubleau, Philippe Blanchard, Simon Richir

Abstract:

Introduction:The widespread availability of artificial intelligence, exemplified by Generative Pre-trained Transformers (GPT) relying on large language models (LLM), has caused a seismic shift in the realm of knowledge. Everyone now has the capacity to swiftly learn how these models can either serve them well or not. Today, conversational AI like ChatGPT is grounded in neural transformer models, a significant advance in natural language processing facilitated by the emergence of renowned LLMs constructed using neural transformer architecture. Inventiveness of an LLM : OpenAI's GPT-3 stands as a premier LLM, capable of handling a broad spectrum of natural language processing tasks without requiring fine-tuning, reliably producing text that reads as if authored by humans. However, even with an understanding of how LLMs respond to questions asked, there may be lurking behind OpenAI’s seemingly endless responses an inventive model yet to be uncovered. There may be some unforeseen reasoning emerging from the interconnection of neural networks here. Just as a Soviet researcher in the 1940s questioned the existence of Common factors in inventions, enabling an Under standing of how and according to what principles humans create them, it is equally legitimate today to explore whether solutions provided by LLMs to complex problems also share common denominators. Theory of Inventive Problem Solving (TRIZ) : We will revisit some fundamentals of TRIZ and how Genrich ALTSHULLER was inspired by the idea that inventions and innovations are essential means to solve societal problems. It's crucial to note that traditional problem-solving methods often fall short in discovering innovative solutions. The design team is frequently hampered by psychological barriers stemming from confinement within a highly specialized knowledge domain that is difficult to question. We presume ChatGPT Utilizes TRIZ 40. Hence, the objective of this research is to decipher the inventive model of LLMs, particularly that of ChatGPT, through a comparative study. This will enhance the efficiency of sustainable innovation processes and shed light on how the construction of a solution to a complex problem was devised. Description of the Experimental Protocol : To confirm or reject our main hypothesis that is to determine whether ChatGPT uses TRIZ, we will follow a stringent protocol that we will detail, drawing on insights from a panel of two TRIZ experts. Conclusion and Future Directions : In this endeavor, we sought to comprehend how an LLM like GPT addresses complex challenges. Our goal was to analyze the inventive model of responses provided by an LLM, specifically ChatGPT, by comparing it to an existing standard model: TRIZ 40. Of course, problem solving is our main focus in our endeavours.

Keywords: artificial intelligence, Triz, ChatGPT, inventiveness, problem-solving

Procedia PDF Downloads 72

Authors: Shreeyam, Ranjan Kumar Sah, Shivangi

Abstract:

Recognizing and solving handwritten mathematical expressions can be a challenging task, particularly when certain characters are segmented and classified. This project proposes a solution that uses Convolutional Neural Network (CNN) and image processing techniques to accurately solve various types of equations, including arithmetic, quadratic, and trigonometric equations, as well as logical operations like logical AND, OR, NOT, NAND, XOR, and NOR. The proposed solution also provides a graphical solution, allowing users to visualize equations and their solutions. In addition to equation solving, the platform, called CNNCalc, offers a comprehensive learning experience for students. It provides educational content, a quiz platform, and a coding platform for practicing programming skills in different languages like C, Python, and Java. This all-in-one solution makes the learning process engaging and enjoyable for students. The proposed methodology includes horizontal compact projection analysis and survey for segmentation and binarization, as well as connected component analysis and integrated connected component analysis for character classification. The compact projection algorithm compresses the horizontal projections to remove noise and obtain a clearer image, contributing to the accuracy of character segmentation. Experimental results demonstrate the effectiveness of the proposed solution in solving a wide range of mathematical equations. CNNCalc provides a powerful and user-friendly platform for solving equations, learning, and practicing programming skills. With its comprehensive features and accurate results, CNNCalc is poised to revolutionize the way students learn and solve mathematical equations. The platform utilizes a custom-designed Convolutional Neural Network (CNN) with image processing techniques to accurately recognize and classify symbols within handwritten equations. The compact projection algorithm effectively removes noise from horizontal projections, leading to clearer images and improved character segmentation. Experimental results demonstrate the accuracy and effectiveness of the proposed solution in solving a wide range of equations, including arithmetic, quadratic, trigonometric, and logical operations. CNNCalc features a user-friendly interface with a graphical representation of equations being solved, making it an interactive and engaging learning experience for users. The platform also includes tutorials, testing capabilities, and programming features in languages such as C, Python, and Java. Users can track their progress and work towards improving their skills. CNNCalc is poised to revolutionize the way students learn and solve mathematical equations with its comprehensive features and accurate results.

Keywords: AL, ML, hand written equation solver, maths, computer, CNNCalc, convolutional neural networks

Procedia PDF Downloads 121

Authors: Niyati Sharma, Om Anand, Sanjeev Kumar Prasad

Abstract:

Assuming future sale value for any of the organizations is one of the major essential characteristics of tactical development. Walmart Sales Forecasting is the finest illustration to work with as a beginner; subsequently, it has the major retail data set. Walmart uses this sales estimate problem for hiring purposes also. We would like to analyzing how the internal and external effects of one of the largest companies in the US can walk out their Weekly Sales in the future. Demand forecasting is the planned prerequisite of products or services in the imminent on the basis of present and previous data and different stages of the market. Since all associations is facing the anonymous future and we do not distinguish in the future good demand. Hence, through exploring former statistics and recent market statistics, we envisage the forthcoming claim and building of individual goods, which are extra challenging in the near future. As a result of this, we are producing the required products in pursuance of the petition of the souk in advance. We will be using several machine learning models to test the exactness and then lastly, train the whole data by Using linear regression and fitting the training data into it. Accuracy is 8.88%. The extra trees regression model gives the best accuracy of 97.15%.

Keywords: random forest algorithm, linear regression algorithm, extra trees classifier, mean absolute error

Procedia PDF Downloads 148

Authors: Radhika Ranjan Roy

Abstract:

Deployment of machine learning (ML)/deep learning (DL) algorithms for cyberattack detection in operational communications networks (wireless and/or wire-line) is being delayed because of low-performance parameters (e.g., recall, precision, and f₁-score). If datasets become imbalanced, which is the usual case for communications networks, the performance tends to become worse. Complexities in handling reducing dimensions of the feature sets for increasing performance are also a huge problem. Mahalanobis algorithms have been widely applied in scientific research because Mahalanobis distance metric learning is a successful framework. In this paper, we have investigated the Mahalanobis binary classifier algorithm for increasing cyberattack detection performance over communications networks as a proof of concept. We have also found that high-dimensional information in intermediate features that are not utilized as much for classification tasks in ML/DL algorithms are the main contributor to the state-of-the-art of improved performance of the Mahalanobis method, even for imbalanced and sparse datasets. With no feature reduction, MD offers uniform results for precision, recall, and f₁-score for unbalanced and sparse NSL-KDD datasets.

Keywords: Mahalanobis distance, machine learning, deep learning, NS-KDD, local intrinsic dimensionality, chi-square, positive semi-definite, area under the curve

Procedia PDF Downloads 77

Authors: Srinaath Anbu Durai, Wang Zhaoxia

Abstract:

Twitter sentiment has been used as a predictor to predict price values or trends in both the stock market and housing market. The pioneering works in this stream of research drew upon works in behavioural economics to show that sentiment or emotions impact economic decisions. Latest works in this stream focus on the algorithm used as opposed to the data used. A literature review of works in this stream through the lens of data used shows that there is a paucity of work that considers the impact of sentiments caused due to an external factor on either the stock or the housing market. This is despite an abundance of works in behavioural economics that show that sentiment or emotions caused due to an external factor impact economic decisions. To address this gap, this research studies the impact of Twitter sentiment pertaining to the Covid-19 pandemic on resale Housing Development Board (HDB) apartment prices in Singapore. It leverages SNSCRAPE to collect tweets pertaining to Covid-19 for sentiment analysis, lexicon based tools VADER and TextBlob are used for sentiment analysis, Granger Causality is used to examine the relationship between Covid-19 cases and the sentiment score, and neural networks are leveraged as prediction models. Twitter sentiment pertaining to Covid-19 as a predictor of HDB price in Singapore is studied in comparison with the traditional predictors of housing prices i.e., the structural and neighbourhood characteristics. The results indicate that using Twitter sentiment pertaining to Covid19 leads to better prediction than using only the traditional predictors and performs better as a predictor compared to two of the traditional predictors. Hence, Twitter sentiment pertaining to an external factor should be considered as important as traditional predictors. This paper demonstrates the real world economic applications of sentiment analysis of Twitter data.

Keywords: sentiment analysis, Covid-19, housing price prediction, tweets, social media, Singapore HDB, behavioral economics, neural networks

Procedia PDF Downloads 113

Authors: Heichia Wang, Yalan Chao

Abstract:

Social inequality is a persistent problem. One of the ways to solve this problem is through education. At present, vulnerable groups are often less geographically accessible to educational resources. However, compared with educational resources, communication equipment is easier for vulnerable groups. Now that information and communication technology (ICT) has entered the field of education, today we can accept the convenience that ICT provides in education, and the mobility that it brings makes learning independent of time and place. With mobile learning, teachers and students can start discussions in an online chat room without the limitations of time or place. However, because liquidity learning is quite convenient, people tend to solve problems in short online texts with lack of detailed information in a lack of convenient online environment to express ideas. Therefore, the ICT education environment may cause misunderstanding between teachers and students. Therefore, in order to better understand each other's views between teachers and students, this study aims to clarify the essays of the analysts and classify the students into several types of learning questions to clarify the views of teachers and students. In addition, this study attempts to extend the description of possible omissions in short texts by using external resources prior to classification. In short, by applying a short text classification, this study can point out each student's learning problems and inform the instructor where the main focus of the future course is, thus improving the ICT education environment. In order to achieve the goals, this research uses convolutional neural network (CNN) method to analyze short discussion content between teachers and students in an ICT education environment. Divide students into several main types of learning problem groups to facilitate answering student problems. In addition, this study will further cluster sub-categories of each major learning type to indicate specific problems for each student. Unlike most neural network programs, this study attempts to extend short texts with external resources before classifying them to improve classification performance. In short, by applying the classification of short texts, we can point out the learning problems of each student and inform the instructors where the main focus of future courses will improve the ICT education environment. The data of the empirical process will be used to pre-process the chat records between teachers and students and the course materials. An action system will be set up to compare the most similar parts of the teaching material with each student's chat history to improve future classification performance. Later, the function of short text classification uses CNN to classify rich chat records into several major learning problems based on theory-driven titles. By applying these modules, this research hopes to clarify the main learning problems of students and inform teachers that they should focus on future teaching.

Keywords: ICT education improvement, social equality, short text analysis, convolutional neural network

Procedia PDF Downloads 127

Authors: Saurabh Kumar

Abstract:

In the fiercely competitive landscape of e-commerce, understanding and mitigating customer churn has become paramount for sustainable business growth. This paper presents a thorough investigation into the application of machine learning techniques for churn prediction in e-commerce, aiming to provide actionable insights for businesses seeking to enhance customer retention strategies. We conduct a comparative study of various machine learning algorithms, including traditional statistical methods and ensemble techniques, leveraging a rich dataset sourced from Kaggle. Through rigorous evaluation, we assess the predictive performance, interpretability, and scalability of each method, elucidating their respective strengths and limitations in capturing the intricate dynamics of customer churn. We identified the XGBoost classifier to be the best performing. Our findings not only offer practical guidelines for selecting suitable modeling approaches but also contribute to the broader understanding of customer behavior in the e-commerce domain. Ultimately, this research equips businesses with the knowledge and tools necessary to proactively identify and address churn, thereby fostering long-term customer relationships and sustaining competitive advantage.

Keywords: customer churn, e-commerce, machine learning techniques, predictive performance, sustainable business growth

Procedia PDF Downloads 26

Authors: Yunwei Tang, Linhai Jing, Hui Li, Qingjie Liu, Xiuxia Li, Qi Yan, Haifeng Ding

Abstract:

The giant panda (Ailuropoda melanoleuca) is an endangered species, mainly live in central China, where bamboos act as the main food source of wild giant pandas. Knowledge of spatial distribution of bamboos therefore becomes important for identifying the habitat of giant pandas. There have been ongoing studies for mapping bamboos and other tree species using remote sensing. WorldView-2 (WV-2) is the first high resolution commercial satellite with eight Multi-Spectral (MS) bands. Recent studies demonstrated that WV-2 imagery has a high potential in classification of tree species. The advanced classification techniques are important for utilising high spatial resolution imagery. It is generally agreed that object-based image analysis is a more desirable method than pixel-based analysis in processing high spatial resolution remotely sensed data. Classifiers that use spatial information combined with spectral information are known as contextual classifiers. It is suggested that contextual classifiers can achieve greater accuracy than non-contextual classifiers. Thus, spatial correlation can be incorporated into classifiers to improve classification results. The study area is located at Wuyipeng area in Wolong, Sichuan Province. The complex environment makes it difficult for information extraction since bamboos are sparsely distributed, mixed with brushes, and covered by other trees. Extensive fieldworks in Wuyingpeng were carried out twice. The first one was on 11th June, 2014, aiming at sampling feature locations for geometric correction and collecting training samples for classification. The second fieldwork was on 11th September, 2014, for the purposes of testing the classification results. In this study, spectral separability analysis was first performed to select appropriate MS bands for classification. Also, the reflectance analysis provided information for expanding sample points under the circumstance of knowing only a few. Then, a spatially weighted object-based k-nearest neighbour (k-NN) classifier was applied to the selected MS bands to identify seven land cover types (bamboo, conifer, broadleaf, mixed forest, brush, bare land, and shadow), accounting for spatial correlation within classes using geostatistical modelling. The spatially weighted k-NN method was compared with three alternatives: the traditional k-NN classifier, the Support Vector Machine (SVM) method and the Classification and Regression Tree (CART). Through field validation, it was proved that the classification result obtained using the spatially weighted k-NN method has the highest overall classification accuracy (77.61%) and Kappa coefficient (0.729); the producer’s accuracy and user’s accuracy achieve 81.25% and 95.12% for the bamboo class, respectively, also higher than the other methods. Photos of tree crowns were taken at sample locations using a fisheye camera, so the canopy density could be estimated. It is found that it is difficult to identify bamboo in the areas with a large canopy density (over 0.70); it is possible to extract bamboos in the areas with a median canopy density (from 0.2 to 0.7) and in a sparse forest (canopy density is less than 0.2). In summary, this study explores the ability of WV-2 imagery for bamboo extraction in a mountainous region in Sichuan. The study successfully identified the bamboo distribution, providing supporting knowledge for assessing the habitats of giant pandas.

Keywords: bamboo mapping, classification, geostatistics, k-NN, worldview-2

Procedia PDF Downloads 311

Authors: Budoor Al Abid

Abstract:

Educational systems are increasingly provided as open online services, providing guidance and support for individual learners. To adapt the learning systems, a proper evaluation must be made. This paper builds the evaluation model Fuzzy C Means Adaptive System (FCMAS) based on data mining techniques to assess the difficulty of the questions. The following steps are implemented; first using a dataset from an online international learning system called (slepemapy.cz) the dataset contains over 1300000 records with 9 features for students, questions and answers information with feedback evaluation. Next, a normalization process as preprocessing step was applied. Then FCM clustering algorithms are used to adaptive the difficulty of the questions. The result is three cluster labeled data depending on the higher Wight (easy, Intermediate, difficult). The FCM algorithm gives a label to all the questions one by one. Then Random Forest (RF) Classifier model is constructed on the clustered dataset uses 70% of the dataset for training and 30% for testing; the result of the model is a 99.9% accuracy rate. This approach improves the Adaptive E-learning system because it depends on the student behavior and gives accurate results in the evaluation process more than the evaluation system that depends on feedback only.

Keywords: machine learning, adaptive, fuzzy logic, data mining

Procedia PDF Downloads 194

Authors: Lavanya Srinivasan

Abstract:

Biometric identification is to identify unique features in a person like fingerprints, iris, ear, and voice recognition that need the subject's permission and physical contact. Gait biometric is used to identify the unique gait of the person by extracting moving features. The main advantage of gait biometric to identify the gait of a person at a distance, without any physical contact. In this work, the gait biometric is used for person re-identification. The person walking naturally compared with the same person walking with bag, coat, and case recorded using longwave infrared, short wave infrared, medium wave infrared, and visible cameras. The videos are recorded in rural and in urban environments. The pre-processing technique includes human identified using YOLO, background subtraction, silhouettes extraction, and synthesis Gait Entropy Image by averaging the silhouettes. The moving features are extracted from the Gait Entropy Energy Image. The extracted features are dimensionality reduced by the principal component analysis and recognised using different classifiers. The comparative results with the different classifier show that linear discriminant analysis outperforms other classifiers with 95.8% for visible in the rural dataset and 94.8% for longwave infrared in the urban dataset.

Keywords: biometric, gait, silhouettes, YOLO

Procedia PDF Downloads 171

Authors: Aileen F. Wang

Abstract:

Mammography has been one of the most reliable methods for early detection and diagnosis of breast cancer. However, mammography misses about 17% and up to 30% of breast cancers due to the subtle and unstable appearances of breast cancer in their early stages. Recent computer-aided diagnosis (CADx) technology using Zernike moments has improved detection accuracy. However, it has several drawbacks: it uses manual segmentation, Zernike moments are not robust, and it still has a relatively high false negative rate (FNR)–17.6%. This project will focus on the development of a novel breast cancer detection algorithm to automatically segment the breast mass and further reduce FNR. The algorithm consists of automatic segmentation of a single breast mass using Point Region Growing Segmentation, reconstruction of the segmented breast mass using Pseudo-Zernike moments, and classification of the breast mass using the root mean square (RMS). A comparative study among the various algorithms on the segmentation and reconstruction of breast masses was performed on randomly selected mammographic images. The results demonstrated that the newly developed algorithm is the best in terms of accuracy and cost effectiveness. More importantly, the new classifier RMS has the lowest FNR–6%.

Keywords: computer aided diagnosis, mammography, point region growing segmentation, pseudo-zernike moments, root mean square

Procedia PDF Downloads 451

Authors: Bliss Singhal

Abstract:

Epilepsy is a prevalent neurological disorder affecting approximately 50 million individuals worldwide and 1.2 million Americans. There exist millions of pediatric patients with intractable epilepsy, a condition in which seizures fail to come under control. The occurrence of seizures can result in physical injury, disorientation, unconsciousness, and additional symptoms that could impede children's ability to participate in everyday tasks. Predicting seizures can help parents and healthcare providers take precautions, prevent risky situations, and mentally prepare children to minimize anxiety and nervousness associated with the uncertainty of a seizure. This research proposes a comprehensive framework to predict seizures in pediatric patients by evaluating machine learning algorithms on unimodal neuroimaging data consisting of electroencephalogram signals. The bandpass filtering and independent component analysis proved to be effective in reducing the noise and artifacts from the dataset. Various machine learning algorithms’ performance is evaluated on important metrics such as accuracy, precision, specificity, sensitivity, F1 score and MCC. The results show that the deep learning algorithms are more successful in predicting seizures than logistic Regression, and k nearest neighbors. The recurrent neural network (RNN) gave the highest precision and F1 Score, long short-term memory (LSTM) outperformed RNN in accuracy and convolutional neural network (CNN) resulted in the highest Specificity. This research has significant implications for healthcare providers in proactively managing seizure occurrence in pediatric patients, potentially transforming clinical practices, and improving pediatric care.

Keywords: intractable epilepsy, seizure, deep learning, prediction, electroencephalogram channels

Procedia PDF Downloads 83

Authors: Ian O'Loughlin

Abstract:

Memory networks aim to integrate some of the recent successes in machine learning with a dynamic memory base that can be updated and deployed in artificial reasoning tasks. These models involve training networks to identify, update, and operate over stored elements in a large memory array in order, for example, to ably perform question and answer tasks parsing real-world and simulated discourses. This family of approaches still faces numerous challenges: the performance of these network models in simulated domains remains considerably better than in open, real-world domains, wide-context cues remain elusive in parsing words and sentences, and even moderately complex sentence structures remain problematic. This innovation, employing an array of stored and updatable ‘memory’ elements over which the system operates as it parses text input and develops responses to questions, is a compelling one for at least two reasons: first, it addresses one of the difficulties that standard machine learning techniques face, by providing a way to store a large bank of facts, offering a way forward for the kinds of long-term reasoning that, for example, recurrent neural networks trained on a corpus have difficulty performing. Second, the addition of a stored long-term memory component in artificial reasoning seems psychologically plausible; human reasoning appears replete with invocations of long-term memory, and the stored but dynamic elements in the arrays of memory networks are deeply reminiscent of the way that human memory is readily and often characterized. However, this apparent psychological plausibility is belied by a recent turn in the study of human memory in cognitive science. In recent years, the very notion that there is a stored element which enables remembering, however dynamic or reconstructive it may be, has come under deep suspicion. In the wake of constructive memory studies, amnesia and impairment studies, and studies of implicit memory—as well as following considerations from the cognitive neuroscience of memory and conceptual analyses from the philosophy of mind and cognitive science—researchers are now rejecting storage and retrieval, even in principle, and instead seeking and developing models of human memory wherein plasticity and dynamics are the rule rather than the exception. In these models, storage is entirely avoided by modeling memory using a recurrent neural network designed to fit a preconceived energy function that attains zero values only for desired memory patterns, so that these patterns are the sole stable equilibrium points in the attractor network. So although the array of long-term memory elements in memory networks seem psychologically appropriate for reasoning systems, they may actually be incurring difficulties that are theoretically analogous to those that older, storage-based models of human memory have demonstrated. The kind of emergent stability found in the attractor network models more closely fits our best understanding of human long-term memory than do the memory network arrays, despite appearances to the contrary.

Keywords: artificial reasoning, human memory, machine learning, neural networks

Procedia PDF Downloads 271

Authors: Adriana Postal, Silvio C. Sampaio, Marcio A. Villas Boas, Josué P. Castro

Abstract:

This study deals with the estimation of reference evapotranspiration (ET₀) in an agricultural context, focusing on efficient irrigation management to meet the growing interest in the sustainable management of water resources. Given the importance of water in agriculture and its scarcity in many regions, efficient use of this resource is essential to ensure food security and environmental sustainability. The methodology used involved the application of artificial intelligence techniques, specifically Multilayer Perceptron (MLP) Artificial Neural Networks (ANNs), to predict ET₀ in the state of Paraná, Brazil. The models were trained and validated with meteorological data from the Brazilian National Institute of Meteorology (INMET), together with data obtained from a producer's weather station in the western region of Paraná. Two optimizers (SGD and Adam) and different meteorological variables, such as temperature, humidity, solar radiation, and wind speed, were explored as inputs to the models. Nineteen configurations with different input variables were tested; amidst them, configuration 9, with 8 input variables, was identified as the most efficient of all. Configuration 10, with 4 input variables, was considered the most effective, considering the smallest number of variables. The main conclusions of this study show that MLP ANNs are capable of accurately estimating ET₀, providing a valuable tool for irrigation management in agriculture. Both configurations (9 and 10) showed promising performance in predicting ET₀. The validation of the models with cultivator data underlined the practical relevance of these tools and confirmed their generalization ability for different field conditions. The results of the statistical metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R²), showed excellent agreement between the model predictions and the observed data, with MAE as low as 0.01 mm/day and 0.03 mm/day, respectively. In addition, the models achieved an R² between 0.99 and 1, indicating a satisfactory fit to the real data. This agreement was also confirmed by the Kolmogorov-Smirnov test, which evaluates the agreement of the predictions with the statistical behavior of the real data and yields values between 0.02 and 0.04 for the producer data. In addition, the results of this study suggest that the developed technique can be applied to other locations by using specific data from these sites to further improve ET₀ predictions and thus contribute to sustainable irrigation management in different agricultural regions. The study has some limitations, such as the use of a single ANN architecture and two optimizers, the validation with data from only one producer, and the possible underestimation of the influence of seasonality and local climate variability. An irrigation management application using the most efficient models from this study is already under development. Future research can explore different ANN architectures and optimization techniques, validate models with data from multiple producers and regions, and investigate the model's response to different seasonal and climatic conditions.

Keywords: agricultural technology, neural networks in agriculture, water efficiency, water use optimization

Procedia PDF Downloads 45

Authors: Hritwik Ghosh, Irfan Sadiq Rahat, Sachi Nandan Mohanty, J. V. R. Ravindra

Abstract:

In the rapidly evolving landscape of medical diagnostics, the early detection and accurate classification of skin cancer remain paramount for effective treatment outcomes. This research delves into the transformative potential of Artificial Intelligence (AI), specifically Deep Learning (DL), as a tool for discerning and categorizing various skin conditions. Utilizing a diverse dataset of 3,000 images representing nine distinct skin conditions, we confront the inherent challenge of class imbalance. This imbalance, where conditions like melanomas are over-represented, is addressed by incorporating class weights during the model training phase, ensuring an equitable representation of all conditions in the learning process. Our pioneering approach introduces a hybrid model, amalgamating the strengths of two renowned Convolutional Neural Networks (CNNs), VGG16 and ResNet50. These networks, pre-trained on the ImageNet dataset, are adept at extracting intricate features from images. By synergizing these models, our research aims to capture a holistic set of features, thereby bolstering classification performance. Preliminary findings underscore the hybrid model's superiority over individual models, showcasing its prowess in feature extraction and classification. Moreover, the research emphasizes the significance of rigorous data pre-processing, including image resizing, color normalization, and segmentation, in ensuring data quality and model reliability. In essence, this study illuminates the promising role of AI and DL in revolutionizing skin cancer diagnostics, offering insights into its potential applications in broader medical domains.

Keywords: artificial intelligence, machine learning, deep learning, skin cancer, dermatology, convolutional neural networks, image classification, computer vision, healthcare technology, cancer detection, medical imaging

Procedia PDF Downloads 85

Authors: Masoud Ghaemi, MohammadJafar Hedayati, Faezeh Yousefzadeh, Hoseinali Heydarzadeh

Abstract:

According to the statistics of the International Committee on Large Dams, internal erosion and piping (scour) are major causes of the destruction of earth-fill dams. If such dams are constructed in narrow valleys, the valley walls will increase the arching of the dam body due to the transfer of vertical and horizontal stresses, so the occurrence of hydraulic fracturing in these embankments is more likely. Roudbar Dam in Lorestan is a clay-core pebble earth-fill dam constructed in a relatively narrow valley in western Iran. Three years after the onset of impoundment, there has been a fall in dam behavior. Evaluation of the dam behavior based on the data recorded on the instruments installed inside the dam body and foundation confirms the occurrence of internal erosion in the lower and adjacent parts of the core on the left support (abutment). The phenomenon of hydraulic fracturing is one of the main causes of the onset of internal erosion in this dam. Accordingly, the main objective of this paper is to evaluate the validity of this hypothesis. To evaluate the validity of this hypothesis, the dam behavior during construction and impoundment has been first simulated with a three-dimensional numerical model. Then, using validated empirical equations, the safety factor of the occurrence of hydraulic fracturing phenomenon upstream of the dam score was calculated. Then, using the artificial neural network, the failure time of the given section was predicted based on the maximum stress trend created. The study results show that steep slopes of valley walls, sudden changes in coefficient, and differences in compressibility properties of dam body materials have caused considerable stress transfer from core to adjacent valley walls, especially at its lower levels. This has resulted in the coefficient of confidence of the occurrence of hydraulic fracturing in each of these areas being close to one in each of the empirical equations used.

Keywords: arching, artificial neural network, FLAC3D, hydraulic fracturing, internal erosion, pore water pressure

Procedia PDF Downloads 175

Authors: Anna Rinaldi, Pierfrancesco Dellino

Abstract:

COVID-19 vaccine hesitancy can be observed at different rates in different countries. In June 2021, 1,068 people were surveyed in France and Italy to inquire about individual potential acceptance, focusing on time preferences in a risk-return framework: having the vaccination today, in a month, and in 3 months; perceived risks of vaccination and COVID-19; and expected benefit of the vaccine. A randomized controlled trial was conducted to understand how everyday stimuli like fact-based news about vaccines impact an audience's acceptance of vaccination. The main experiment involved two groups of participants and two different articles about vaccine-related thrombosis taken from two Italian newspapers. One article used a more abstract description and language, and the other used a more anecdotal description and concrete language; each group read only one of these articles. Two other groups were assigned categorization tasks; one was asked to complete a concrete categorization task, and the other an abstract categorization task. Individual preferences for vaccination were found to be variable and unstable over time, and individual choices of accepting, refusing, or delaying could be affected by the way news is written. In order to understand these dynamic preferences, the present work proposes a new model based on seven categories of human behaviors that were validated by a neural network. A treatment effect was observed: participants who read the articles shifted to vaccine hesitancy categories more than participants assigned to other treatments and control. Furthermore, there was a significant gender effect, showing that the type of language leading to a lower hesitancy rate for men is correlated with a higher hesitancy rate for women and vice versa. This outcome should be taken into consideration for an appropriate gender-based communication campaign aimed at achieving herd immunity. The trial was registered at ClinicalTrials.gov NCT05582564 (17/10/2022).

Keywords: vaccine hesitancy, risk elicitation, neural network, covid19

Procedia PDF Downloads 82

Authors: Medha Basu, Kumardeb Banerjee, Dipak Ghosh

Abstract:

After the massive blow of the COVID-19 pandemic, several health hazards have been reported all over the world. Serious cases of Major Depressive Disorder (MDD) are seen to be common in about 15% of the global population, making depression one of the leading mental health diseases, as reported by the World Health Organization. Various psychological and pharmacological treatment techniques are regularly being reported. Music, a globally accepted mode of entertainment, is often used as a therapeutic measure to treat various health conditions. We have tried to understand how Indian Classical Music can affect the overall well-being of the human brain. A case study has been reported here, where a Flute-rendition has been chosen from a detailed audience response survey, and the effects of that clip on human brain conditions have been studied from both psychological and neural perspectives. Taking help from internationally-accepted depression-rating scales, two questionnaires have been designed to understand both the prolonged and immediate effect of music on various emotional states of human lives. Thereafter, from EEG experiments on 5 participants using the same clip, the parameter ‘ALAY’, alpha frontal asymmetry (alpha power difference of right and left frontal hemispheres), has been calculated. Works of Richard Davidson show that an increase in the ‘ALAY’ value indicates a decrease in depressive symptoms. Using the non-linear technique of MFDFA on EEG analysis, we have also calculated frontal asymmetry using the complexity values of alpha-waves in both hemispheres. The results show a positive correlation between both the psychological survey and the EEG findings, revealing the prominent role of music on the human brain, leading to a decrease in mental unrest and an increase in overall well-being. In this study, we plan to propose the scientific foundation of music therapy, especially from a neurocognition perspective, with appropriate neural bio-markers to understand the positive and remedial effects of music on the human brain.

Keywords: music therapy, EEG, psychological survey, frontal alpha asymmetry, wellbeing

Procedia PDF Downloads 40

Authors: S. Sabatino, V. Calderaro, V. Galdi, G. Graber, L. Ippolito

Abstract:

Climate change is a rapidly growing global threat caused mainly by increased emissions of carbon dioxide (CO₂) into the atmosphere. These emissions come from multiple sources, including industry, power generation, and the transport sector. The need to tackle climate change and reduce CO₂ emissions is indisputable. A crucial solution to achieving decarbonization in the transport sector is the adoption of electric vehicles (EVs). These vehicles use lithium (Li-Ion) batteries as an energy source, making them extremely efficient and with low direct emissions. However, Li-Ion batteries are not without problems, including the risk of overheating and performance degradation. To ensure its safety and longevity, it is essential to use a battery management system (BMS). The BMS constantly monitors battery status, adjusts temperature and cell balance, ensuring optimal performance and preventing dangerous situations. From the monitoring carried out, it is also able to optimally manage the battery to increase its life. Among the parameters monitored by the BMS, the main ones are State of Charge (SoC), State of Health (SoH), and State of Power (SoP). The evaluation of these parameters can be carried out in two ways: offline, using benchtop batteries tested in the laboratory, or online, using batteries installed in moving vehicles. Online estimation is the preferred approach, as it relies on capturing real-time data from batteries while operating in real-life situations, such as in everyday EV use. Actual battery usage conditions are highly variable. Moving vehicles are exposed to a wide range of factors, including temperature variations, different driving styles, and complex charge/discharge cycles. This variability is difficult to replicate in a controlled laboratory environment and can greatly affect performance and battery life. Online estimation captures this variety of conditions, providing a more accurate assessment of battery behavior in real-world situations. In this article, a hybrid approach based on a neural network and a statistical method for real-time estimation of SoC, SoH, and SoP parameters of interest is proposed. These parameters are estimated from the analysis of a one-day driving profile of an electric vehicle, assumed to be divided into the following four phases: (i) Partial discharge (SoC 100% - SoC 50%), (ii) Partial discharge (SoC 50% - SoC 80%), (iii) Deep Discharge (SoC 80% - SoC 30%) (iv) Full charge (SoC 30% - SoC 100%). The neural network predicts the values of ohmic resistance and incremental capacity, while the statistical method is used to estimate the parameters of interest. This reduces the complexity of the model and improves its prediction accuracy. The effectiveness of the proposed model is evaluated by analyzing its performance in terms of square mean error (RMSE) and percentage error (MAPE) and comparing it with the reference method found in the literature.

Keywords: electric vehicle, Li-Ion battery, BMS, state-of-charge, state-of-health, state-of-power, artificial neural networks

Procedia PDF Downloads 67

Authors: Priyadarsini Samal, Rajesh Singla

Abstract:

Many mobile games come with the benefits of entertainment by introducing stress to the human brain. In recognizing this mental stress, the brain-computer interface (BCI) plays an important role. It has various neuroimaging approaches which help in analyzing the brain signals. Electroencephalogram (EEG) is the most commonly used method among them as it is non-invasive, portable, and economical. Here, this paper investigates the pattern in brain signals when introduced with mental stress. Two healthy volunteers played a game whose aim was to search hidden words from the grid, and the levels were chosen randomly. The EEG signals during gameplay were recorded to investigate the impacts of stress with the changing levels from easy to medium to hard. A total of 16 features of EEG were analyzed for this experiment which includes power band features with relative powers, event-related desynchronization, along statistical features. Support vector machine was used as the classifier, which resulted in an accuracy of 93.9% for three-level stress analysis; for two levels, the accuracy of 92% and 98% are achieved. In addition to that, another game that was similar in nature was played by the volunteers. A suitable regression model was designed for prediction where the feature sets of the first and second game were used for testing and training purposes, respectively, and an accuracy of 73% was found.

Keywords: brain computer interface, electroencephalogram, regression model, stress, word search

Procedia PDF Downloads 185

Authors: Ben Khayut, Lina Fabri, Maya Avikhana

Abstract:

The models of the modern Artificial Narrow Intelligence (ANI) cannot: a) independently and continuously function without of human intelligence, used for retraining and reprogramming the ANI’s models, and b) think, understand, be conscious, cognize, infer, and more in state of Uncertainty, and changes in situations, and environmental objects. To eliminate these shortcomings and build a new generation of Artificial Intelligence systems, the paper proposes a Conception, Model, and Method of Computational Psycholinguistic Cognitive Situational-Fuzzy Self-Controlled Brain and Mind System (CPCSFSCBMSUU) using a neural network as its computational memory, operating under uncertainty, and activating its functions by perception, identification of real objects, fuzzy situational control, forming images of these objects, modeling their psychological, linguistic, cognitive, and neural values of properties and features, the meanings of which are identified, interpreted, generated, and formed taking into account the identified subject area, using the data, information, knowledge, and images, accumulated in the Memory. The functioning of the CPCSFSCBMSUU is carried out by its subsystems of the: fuzzy situational control of all processes, computational perception, identifying of reactions and actions, Psycholinguistic Cognitive Fuzzy Logical Inference, Decision making, Reasoning, Systems Thinking, Planning, Awareness, Consciousness, Cognition, Intuition, Wisdom, analysis and processing of the psycholinguistic, subject, visual, signal, sound and other objects, accumulation and using the data, information and knowledge in the Memory, communication, and interaction with other computing systems, robots and humans in order of solving the joint tasks. To investigate the functional processes of the proposed system, the principles of Situational Control, Fuzzy Logic, Psycholinguistics, Informatics, and modern possibilities of Data Science were applied. The proposed self-controlled System of Brain and Mind is oriented on use as a plug-in in multilingual subject Applications.

Keywords: computational brain, mind, psycholinguistic, system, under uncertainty

Procedia PDF Downloads 176

Authors: Luis Alvarado, Victor Poblete, Isaac Gonzalez, Yetzabeth Gonzalez

Abstract:

In MIREX 2016 (http://www.music-ir.org/mirex), the deep neural network (DNN)-Deep Chroma Extractor, proposed by Korzeniowski and Wiedmer, reached the highest score in an audio chord recognition task. In the present paper, this tool is assessed under acoustic reverberant environments and distinct source-microphone distances. The evaluation dataset comprises The Beatles and Queen datasets. These datasets are sequentially re-recorded with a single microphone in a real reverberant chamber at four reverberation times (0 -anechoic-, 1, 2, and 3 s, approximately), as well as four source-microphone distances (32, 64, 128, and 256 cm). It is expected that the performance of the trained DNN will dramatically decrease under these acoustic conditions with signals degraded by room reverberation and distance to the source. Recently, the effect of the bio-inspired Locally-Normalized Cepstral Coefficients (LNCC), has been assessed in a text independent speaker verification task using speech signals degraded by additive noise at different signal-to-noise ratios with variations of recording distance, and it has also been assessed under reverberant conditions with variations of recording distance. LNCC showed a performance so high as the state-of-the-art Mel Frequency Cepstral Coefficient filters. Based on these results, this paper proposes a variation of locally-normalized triangular filters called Locally-Normalized Quarter Tone (LNQT) filters. By using the LNQT spectrogram, robustness improvements of the trained Deep Chroma Extractor are expected, compared with classical triangular filters, and thus compensating the music signal degradation improving the accuracy of the chord recognition system.

Keywords: chord recognition, deep neural networks, feature extraction, music information retrieval

Procedia PDF Downloads 231

Authors: Han-Jeong Hwang, Jae-Hyun Jo, Fatemeh Alimardani

Abstract:

An accurate diagnosis of psychiatric diseases is a challenging issue, in particular when distinct symptoms for different diseases are overlapped, such as delusions appeared in bipolar mood disorder (BMD) and schizophrenia (SCH). In the present study, we propose a useful way to discriminate BMD and SCH using electroencephalography (EEG). A total of thirty BMD and SCH patients (15 vs. 15) took part in our experiment. EEG signals were measured with nineteen electrodes attached on the scalp using the international 10-20 system, while they were exposed to a visual stimulus flickering at 16 Hz for 95 s. The flickering visual stimulus induces a certain brain signal, known as steady-state visual evoked potential (SSVEP), which is differently observed in patients with BMD and SCH, respectively, in terms of SSVEP amplitude because they process the same visual information in own unique way. For classifying BDM and SCH patients, machine learning technique was employed in which leave-one-out-cross validation was performed. The SSVEPs induced at the fundamental (16 Hz) and second harmonic (32 Hz) stimulation frequencies were extracted using fast Fourier transformation (FFT), and they were used as features. The most discriminative feature was selected using the Fisher score, and support vector machine (SVM) was used as a classifier. From the analysis, we could obtain a classification accuracy of 83.33 %, showing the feasibility of discriminating patients with BMD and SCH using EEG. We expect that our approach can be utilized for psychiatrists to more accurately diagnose the psychiatric disorders, BMD and SCH.

Keywords: bipolar mood disorder, electroencephalography, schizophrenia, machine learning

Procedia PDF Downloads 418

Authors: Jinyan Li, Simon Fong, Raymond Wong, Mohammed Sabah, Fiaidhi Jinan

Abstract:

Clinical data analysis and forecasting have make great contributions to disease control, prevention and detection. However, such data usually suffer from highly unbalanced samples in class distributions. In this paper, we target at the binary imbalanced dataset, where the positive samples take up only the minority. We investigate two different meta-heuristic algorithms, particle swarm optimization and bat-inspired algorithm, and combine both of them with the synthetic minority over-sampling technique (SMOTE) for processing the datasets. One approach is to process the full dataset as a whole. The other is to split up the dataset and adaptively process it one segment at a time. The experimental results reveal that while the performance improvements obtained by the former methods are not scalable to larger data scales, the later one, which we call Adaptive Swarm Balancing Algorithms, leads to significant efficiency and effectiveness improvements on large datasets. We also find it more consistent with the practice of the typical large imbalanced medical datasets. We further use the meta-heuristic algorithms to optimize two key parameters of SMOTE. Leading to more credible performances of the classifier, and shortening the running time compared with the brute-force method.

Keywords: Imbalanced dataset, meta-heuristic algorithm, SMOTE, big data

Procedia PDF Downloads 439

Authors: Andrew N. Saylor, James R. Peters

Abstract:

Scoliosis is a complex 3D deformity of the thoracic and lumbar spines, clinically diagnosed by measurement of a Cobb angle of 10 degrees or more on a coronal X-ray. The Cobb angle is the angle made by the lines drawn along the proximal and distal endplates of the respective proximal and distal vertebrae comprising the curve. Traditionally, Cobb angles are measured manually using either a marker, straight edge, and protractor or image measurement software. The task of measuring the Cobb angle can also be represented by a function taking the spine geometry rendered using X-ray imaging as input and returning the approximate angle. Although the form of such a function may be unknown, it can be approximated using artificial neural networks (ANNs). The performance of ANNs is affected by many factors, including the choice of activation function and network architecture; however, the effects of these parameters on the accuracy of scoliotic deformity measurements are poorly understood. Therefore, the objective of this study was to systematically investigate the effect of ANN architecture and activation function on Cobb angle measurement from the coronal X-rays of scoliotic subjects. The data set for this study consisted of 609 coronal chest X-rays of scoliotic subjects divided into 481 training images and 128 test images. These data, which included labeled Cobb angle measurements, were obtained from the SpineWeb online database. In order to normalize the input data, each image was resized using bi-linear interpolation to a size of 500 × 187 pixels, and the pixel intensities were scaled to be between 0 and 1. A fully connected (dense) ANN with a fixed cost function (mean squared error), batch size (10), and learning rate (0.01) was developed using Python Version 3.7.3 and TensorFlow 1.13.1. The activation functions (sigmoid, hyperbolic tangent [tanh], or rectified linear units [ReLU]), number of hidden layers (1, 3, 5, or 10), and number of neurons per layer (10, 100, or 1000) were varied systematically to generate a total of 36 network conditions. Stochastic gradient descent with early stopping was used to train each network. Three trials were run per condition, and the final mean squared errors and mean absolute errors were averaged to quantify the network response for each condition. The network that performed the best used ReLU neurons had three hidden layers, and 100 neurons per layer. The average mean squared error of this network was 222.28 ± 30 degrees2, and the average mean absolute error was 11.96 ± 0.64 degrees. It is also notable that while most of the networks performed similarly, the networks using ReLU neurons, 10 hidden layers, and 1000 neurons per layer, and those using Tanh neurons, one hidden layer, and 10 neurons per layer performed markedly worse with average mean squared errors greater than 400 degrees2 and average mean absolute errors greater than 16 degrees. From the results of this study, it can be seen that the choice of ANN architecture and activation function has a clear impact on Cobb angle inference from coronal X-rays of scoliotic subjects.

Keywords: scoliosis, artificial neural networks, cobb angle, medical imaging

Procedia PDF Downloads 129

Authors: Yunbin Deng

Abstract:

Most existing applications of automatic speech recognition relies on cooperative subjects at a short distance to a microphone. Standoff speech recognition using microphone arrays can extend the subject to sensor distance somewhat, but it is still limited to only a few feet. As such, most deployed applications of standoff speech recognitions are limited to indoor use at short range. Moreover, these applications require air passway between the subject and the sensor to achieve reasonable signal to noise ratio. This study reports long range (50 feet) automatic speech recognition experiments using a Laser Doppler Vibrometer (LDV) sensor. This study shows that the LDV sensor modality can extend the speech acquisition standoff distance far beyond microphone arrays to hundreds of feet. In addition, LDV enables 'listening' through the windows for uncooperative subjects. This enables new capabilities in automatic audio and speech intelligence, surveillance, and reconnaissance (ISR) for law enforcement, homeland security and counter terrorism applications. The Polytec LDV model OFV-505 is used in this study. To investigate the impact of different vibrating materials, five parallel LDV speech corpora, each consisting of 630 speakers, are collected from the vibrations of a glass window, a metal plate, a plastic box, a wood slate, and a concrete wall. These are the common materials the application could encounter in a daily life. These data were compared with the microphone counterpart to manifest the impact of various materials on the spectrum of the LDV speech signal. State of the art deep neural network modeling approaches is used to conduct continuous speaker independent speech recognition on these LDV speech datasets. Preliminary phoneme recognition results using time-delay neural network, bi-directional long short term memory, and model fusion shows great promise of using LDV for long range speech recognition. To author’s best knowledge, this is the first time an LDV is reported for long distance speech recognition application.

Keywords: covert speech acquisition, distant speech recognition, DSR, laser Doppler vibrometer, LDV, speech intelligence surveillance and reconnaissance, ISR

Procedia PDF Downloads 177