Search results for: speaker recognition

Authors: Yalong Jiang, Zheru Chi

Abstract:

In this paper, we study the factors which determine the capacity of a Convolutional Neural Network (CNN) model and propose the ways to evaluate and adjust the capacity of a CNN model for best matching to a specific pattern recognition task. Firstly, a scheme is proposed to adjust the number of independent functional units within a CNN model to make it be better fitted to a task. Secondly, the number of independent functional units in the capsule network is adjusted to fit it to the training dataset. Thirdly, a method based on Bayesian GAN is proposed to enrich the variances in the current dataset to increase its complexity. Experimental results on the PASCAL VOC 2010 Person Part dataset and the MNIST dataset show that, in both conventional CNN models and capsule networks, the number of independent functional units is an important factor that determines the capacity of a network model. By adjusting the number of functional units, the capacity of a model can better match the complexity of a dataset.

Keywords: CNN, convolutional neural network, capsule network, capacity optimization, character recognition, data augmentation, semantic segmentation

Procedia PDF Downloads 116

Authors: Xiaoxiao Li, Shuangcheng Jia, Qian Li

Abstract:

Keypoint detection has always been a challenge in the field of image recognition. This paper proposes a novelty keypoint detection method which is called Multi-Scale Feature Fusion Convolutional Network with Attention (MFFCNA). We verified that the multi-scale features with the attention mechanism module have better feature expression capability. The feature fusion between different scales makes the information that the network model can express more abundant, and the network is easier to converge. On our self-made street sign corner dataset, we validate the MFFCNA model with an accuracy of 97.8% and a recall of 81%, which are 5 and 8 percentage points higher than the HRNet network, respectively. On the COCO dataset, the AP is 71.9%, and the AR is 75.3%, which are 3 points and 2 points higher than HRNet, respectively. Extensive experiments show that our method has a remarkable improvement in the keypoint recognition tasks, and the recognition effect is better than the existing methods. Moreover, our method can be applied not only to keypoint detection but also to image classification and semantic segmentation with good generality.

Keywords: keypoint detection, feature fusion, attention, semantic segmentation

Procedia PDF Downloads 89

Authors: Khadijat T. Bamigbade, Olufade F. W. Onifade

Abstract:

The field of automatic facial expression analysis has been an active research area in the last two decades. Its vast applicability in various domains has drawn so much attention into developing techniques and dataset that mirror real life scenarios. Many techniques such as Local Binary Patterns and its variants (CLBP, LBP-TOP) and lately, deep learning techniques, have been used for facial expression recognition. However, the problem of occlusion has not been sufficiently handled, making their results not applicable in real life situations. This paper develops a simple, yet highly efficient method tagged Local Binary Pattern-Histogram of Gradient (LBP-HOG) with occlusion detection in face image, using a multi-class SVM for Action Unit and in turn expression recognition. Our method was evaluated on three publicly available datasets which are JAFFE, CK, SFEW. Experimental results showed that our approach performed considerably well when compared with state-of-the-art algorithms and gave insight to occlusion detection as a key step to handling expression in wild.

Keywords: automatic facial expression analysis, local binary pattern, LBP-HOG, occlusion detection

Procedia PDF Downloads 130

Authors: Nassib Abdallah, Pierre Chauvet, Abd El Salam Hajjar, Bassam Daya

Abstract:

In this paper, we propose an optimized brain computer interface (BCI) system for unspoken speech recognition, based on the fact that the constructions of unspoken words rely strongly on the Wernicke area, situated in the temporal lobe. Our BCI system has four modules: (i) the EEG Acquisition module based on a non-invasive headset with 14 electrodes; (ii) the Preprocessing module to remove noise and artifacts, using the Common Average Reference method; (iii) the Features Extraction module, using Wavelet Packet Transform (WPT); (iv) the Classification module based on a one-hidden layer artificial neural network. The present study consists of comparing the recognition accuracy of 5 Arabic words, when using all the headset electrodes or only the 4 electrodes situated near the Wernicke area, as well as the selection effect of the subbands produced by the WPT module. After applying the articial neural network on the produced database, we obtain, on the test dataset, an accuracy of 83.4% with all the electrodes and all the subbands of 8 levels of the WPT decomposition. However, by using only the 4 electrodes near Wernicke Area and the 6 middle subbands of the WPT, we obtain a high reduction of the dataset size, equal to approximately 19% of the total dataset, with 67.5% of accuracy rate. This reduction appears particularly important to improve the design of a low cost and simple to use BCI, trained for several words.

Keywords: brain-computer interface, speech recognition, artificial neural network, electroencephalography, EEG, wernicke area

Procedia PDF Downloads 241

Authors: Youhang Zhou, Weimin Zeng, Qi Xie

Abstract:

Abstract: In order to solve the wear recognition problem of the machine tool guide surface, a new machine tool guide surface recognition method based on the radar-graph barycentre feature is presented in this paper. Firstly, the gray mean value, skewness, projection variance, flat degrees and kurtosis features of the guide surface image data are defined as primary characteristics. Secondly, data Visualization technology based on radar graph is used. The visual barycentre graphical feature is demonstrated based on the radar plot of multi-dimensional data. Thirdly, a classifier based on the support vector machine technology is used, the radar-graph barycentre feature and wear original feature are put into the classifier separately for classification and comparative analysis of classification and experiment results. The calculation and experimental results show that the method based on the radar-graph barycentre feature can detect the guide surface effectively.

Keywords: guide surface, wear defects, feature extraction, data visualization

Procedia PDF Downloads 472

Authors: Monday Agbonica Bello Eje

Abstract:

The language of the print media is characterized by the use of apposition. This linguistic element function strategically in journalistic discourse where it is communicatively necessary to name individuals and provide information about them. Linguistic studies on the language of the print media with bias for apposition have largely dwelt on other areas but the examination of the typology of appositive reference in newspaper discourse. Yet, it is capable of revealing ways writers communicate and provide information necessary for readers to follow and understand the message. The study, therefore, analyses the patterns of appositional occurrences and the typology of reference in newspaper articles. The data were obtained from The Punch and Daily Trust Newspapers. A total of six editions of these newspapers were collected randomly spread over three months. News and feature articles were used in the analysis. Guided by the referential theory of meaning in discourse, the appositions identified were subjected to analysis. The findings show that the semantic relation of coreference and speaker coreference have the highest percentage and frequency of occurrence in the data. This is because the subject matter of news reports and feature articles focuses on humans and the events around them; as a result, readers need to be provided with some form of detail and background information in order to identify as well as follow the discourse. Also, the non-referential relation of absolute synonymy and speaker synonymy no doubt have fewer occurrences and percentages in the analysis. This is tied to a major feature of the language of the media: simplicity. The paper concludes that appositions is mainly used for the purpose of providing the reader with much detail. In this way, the writer transmits information which helps him not only to give detailed yet concise descriptions but also in some way help the reader to follow the discourse.

Keywords: apposition, discourse, newspaper, Nigeria, reference

Procedia PDF Downloads 129

Authors: Ria Ghosh, Soumendra Singh, Dipanjan Mukherjee, Susmita Mondal, Monojit Das, Uttam Pal, Aniruddha Adhikari, Aman Bhushan, Surajit Bose, Siddharth Sankar Bhattacharyya, Debasish Pal, Tanusri Saha-Dasgupta, Maitree Bhattacharyya, Debasis Bhattacharyya, Asim Kumar Mallick, Ranjan Das, Samir Kumar Pal

Abstract:

Drug delivery to a target without adverse effects is one of the major criteria for clinical use. Herein, we have made an attempt to explore the delivery efficacy of SDS surfactant in a monomer and micellar stage during the delivery of the model drug, Toluidine Blue (TB) from the micellar cavity to DNA. Molecular recognition of pre-micellar SDS encapsulated TB with DNA occurs at a rate constant of k1 ~652 s 1. However, no significant release of encapsulated TB at micellar concentration was observed within the experimental time frame. This originated from the higher binding affinity of TB towards the nano-cavity of SDS at micellar concentration which does not allow the delivery of TB from the nano-cavity of SDS micelles to DNA. Thus, molecular recognition controls the extent of DNA recognition by TB which in turn modulates the rate of delivery of TB from SDS in a concentration-dependent manner.

Keywords: DNA, drug delivery, micelle, pre-micelle, SDS, toluidine blue

Procedia PDF Downloads 65

Authors: Uswah Khairuddin, Rubiyah Yusof, Nenny Ruthfalydia Rosli

Abstract:

This paper presents a comparative study between k-Nearest Neighbour (k-NN) and Multi-Layer Perceptron Neural Network (MLP-NN) classifier using Genetic Algorithm (GA) as feature selector for wood recognition system. The features have been extracted from the images using Grey Level Co-Occurrence Matrix (GLCM). The use of GA based feature selection is mainly to ensure that the database used for training the features for the wood species pattern classifier consists of only optimized features. The feature selection process is aimed at selecting only the most discriminating features of the wood species to reduce the confusion for the pattern classifier. This feature selection approach maintains the ‘good’ features that minimizes the inter-class distance and maximizes the intra-class distance. Wrapper GA is used with k-NN classifier as fitness evaluator (GA-kNN). The results shows that k-NN is the best choice of classifier because it uses a very simple distance calculation algorithm and classification tasks can be done in a short time with good classification accuracy.

Keywords: feature selection, genetic algorithm, optimization, wood recognition system

Procedia PDF Downloads 502

Authors: A. Shoiynbek, K. Kozhakhmet, P. Menezes, D. Kuanyshbay, D. Bayazitov

Abstract:

Speech emotion recognition has received increasing research interest all through current years. There was used emotional speech that was collected under controlled conditions in most research work. Actors imitating and artificially producing emotions in front of a microphone noted those records. There are four issues related to that approach, namely, (1) emotions are not natural, and it means that machines are learning to recognize fake emotions. (2) Emotions are very limited by quantity and poor in their variety of speaking. (3) There is language dependency on SER. (4) Consequently, each time when researchers want to start work with SER, they need to find a good emotional database on their language. In this paper, we propose the approach to create an automatic tool for speech emotion extraction based on facial emotion recognition and describe the sequence of actions of the proposed approach. One of the first objectives of the sequence of actions is a speech detection issue. The paper gives a detailed description of the speech detection model based on a fully connected deep neural network for Kazakh and Russian languages. Despite the high results in speech detection for Kazakh and Russian, the described process is suitable for any language. To illustrate the working capacity of the developed model, we have performed an analysis of speech detection and extraction from real tasks.

Keywords: deep neural networks, speech detection, speech emotion recognition, Mel-frequency cepstrum coefficients, collecting speech emotion corpus, collecting speech emotion dataset, Kazakh speech dataset

Procedia PDF Downloads 66

Authors: Mais Nijim, Rama Devi Chennuboyina, Waseem Al Aqqad

Abstract:

Advances in spatial and spectral resolution of satellite images have led to tremendous growth in large image databases. The data we acquire through satellites, radars and sensors consists of important geographical information that can be used for remote sensing applications such as region planning, disaster management. Spatial data classification and object recognition are important tasks for many applications. However, classifying objects and identifying them manually from images is a difficult task. Object recognition is often considered as a classification problem, this task can be performed using machine-learning techniques. Despite of many machine-learning algorithms, the classification is done using supervised classifiers such as Support Vector Machines (SVM) as the area of interest is known. We proposed a classification method, which considers neighboring pixels in a region for feature extraction and it evaluates classifications precisely according to neighboring classes for semantic interpretation of region of interest (ROI). A dataset has been created for training and testing purpose; we generated the attributes by considering pixel intensity values and mean values of reflectance. We demonstrated the benefits of using knowledge discovery and data-mining techniques, which can be on image data for accurate information extraction and classification from high spatial resolution remote sensing imagery.

Keywords: remote sensing, object recognition, classification, data mining, waterbody identification, feature extraction

Procedia PDF Downloads 309

Authors: Nkereke M. Essien

Abstract:

The intention of this paper is to examine stress in the Educated Nigerian Accent of English (ENAE) with the aim of analyzing stress and rhythmic patterns of Nigerian English. Our aim also is to isolate differences and similarities in the stress patterns studied and also know what forms the accent of these Educated Nigerian English (ENE) which marks them off from other groups or English’s of the world, to ascertain and characterize it and to provide documented evidence for its existence. Nigerian stress and rhythmic patterns are significantly different from the British English stress and rhythmic patterns consequently, the educated Nigerian English (ENE) features more stressed syllables than the native speakers’ varieties. The excessive stressed of syllables causes a contiguous “Ss” in the rhythmic flow of ENE, and this brings about a “jerky rhythm’ which distorts communication. To ascertain this claim, ten (10) Nigerian speakers who are educated in the English Language were selected by a stratified Random Sampling technique from two Federal Universities in Nigeria. This classification belongs to the education to the educated class or standard variety. Their performance was compared to that of a Briton (control). The Metrical system of analysis was used. The respondents were made to read some words and utterance which was recorded and analyzed perceptually, statistically and acoustically using the one-way Analysis of Variance (ANOVA). The Turky-Kramer Post Hoc test, the Wilcoxon Matched Pairs Signed Ranks test, and the Praat analysis software were used in the analysis. It was revealed from our findings that the Educated Nigerian English speakers feature more stressed syllables in their productions by spending more time in pronouncing stressed syllables and sometimes lesser time in pronouncing the unstressed syllables. Their overall tempo was faster. The ENE speakers used tone to mark prominence while the native speaker used stress to mark pronounce, typified by the control. We concluded that the stress pattern of the ENE speakers was significantly different from the native speaker’s variety represented by the control’s performance.

Keywords: accent, Nigerian English, rhythm, stress

Procedia PDF Downloads 192

Authors: Megan Weisbart

Abstract:

Background/Significance: The Intensive Care Unit (ICU) environment contributes to nurse burnout. Burnout costs include decreased employee compassion, missed workdays, worse patient outcomes, diminished job performance, high turnover, and higher organizational cost. Meaningful recognition, nurturing of interpersonal connections, and mindfulness-based interventions are associated with decreased burnout. The purpose of this quality improvement project was to decrease Neonatal ICU (NICU) nurse burnout using a Wellness Bundle that fosters meaningful recognition, interpersonal connections and includes mindfulness-based interventions. Methods: The Professional Quality of Life Scale Version 5 (ProQOL5) was used to measure burnout before Wellness Bundle implementation, after six months, and will be given yearly for three years. Meaningful recognition bundle items include Online submission and posting of staff shoutouts, recognition events, Nurses Week and Unit Practice Council member gifts, and an employee recognition program. Fostering of interpersonal connections bundle items include: Monthly staff games with prizes, social events, raffle fundraisers, unit blog, unit wellness basket, and a wellness resource sheet. Quick coherence techniques were implemented at staff meetings and huddles as a mindfulness-based intervention. Findings: The mean baseline burnout score of 14 NICU nurses was 20.71 (low burnout). The baseline range was 13-28, with 11 nurses experiencing low burnout, three nurses experiencing moderate burnout, and zero nurses experiencing high burnout. After six months of the Wellness Bundle Implementation, the mean burnout score of 39 NICU nurses was 22.28 (low burnout). The range was 14-31, with 22 nurses experiencing low burnout, 17 nurses experiencing moderate burnout, and zero nurses experiencing high burnout. Conclusion: A NICU Wellness Bundle that incorporated meaningful recognition, fostering of interpersonal connections, and mindfulness-based activities was implemented to improve work environments and decrease nurse burnout. Participation bias and low baseline response rate may have affected the reliability of the data and necessitate another comparative measure of burnout in one year.

Keywords: burnout, NICU, nurse, wellness

Procedia PDF Downloads 54

Authors: Umair Rashid

Abstract:

Normally softmax loss is used as the supervision signal in face recognition (FR) system, and it boosts the separability of features. In the last two years, a number of techniques have been proposed by reformulating the original softmax loss to enhance the discriminating power of Deep Convolutional Neural Networks (DCNNs) for FR system. To learn angularly discriminative features Cosine-Margin based softmax has been adjusted as monotonically decreasing angular function, that is the main challenge for angular based softmax. On that issue, we propose monotonically decreasing element for Cosine-Margin based softmax and also, we discussed the effect of different monotonically decreasing parameters on angular Margin softmax for FR system. We train the model on publicly available dataset CASIA- WebFace via our proposed monotonically decreasing parameters for cosine function and the tests on YouTube Faces (YTF, Labeled Face in the Wild (LFW), VGGFace1 and VGGFace2 attain the state-of-the-art performance.

Keywords: deep convolutional neural networks, cosine margin face recognition, softmax loss, monotonically decreasing parameter

Procedia PDF Downloads 55

Authors: Subir Gupta, Subhas Ganguly

Abstract:

In this paper, we demonstrate the new area of application of image processing in metallurgical images to develop the more opportunity for structure-property correlation based approaches of alloy design. The present exercise focuses on the development of image processing tools suitable for phrase segmentation, grain boundary detection and recognition of micro-constituents in SEM micrographs of ferrite and pearlite steels. A comprehensive data of micrographs have been experimentally developed encompassing the variation of ferrite and pearlite volume fractions and taking images at different magnification (500X, 1000X, 15000X, 2000X, 3000X and 5000X) under scanning electron microscope. The variation in the volume fraction has been achieved using four different plain carbon steel containing 0.1, 0.22, 0.35 and 0.48 wt% C heat treated under annealing and normalizing treatments. The obtained data pool of micrographs arbitrarily divided into two parts to developing training and testing sets of micrographs. The statistical recognition features for ferrite and pearlite constituents have been developed by learning from training set of micrographs. The obtained features for microstructure pattern recognition are applied to test set of micrographs. The analysis of the result shows that the developed strategy can successfully detect the micro constitutes across the wide range of magnification and variation of volume fractions of the constituents in the structure with an accuracy of about +/- 5%.

Keywords: SEM micrograph, metallurgical image processing, ferrite pearlite steel, microstructure

Procedia PDF Downloads 167

Authors: Marius Zemp

Abstract:

Evidentials indicate how the speaker obtained the information conveyed in a statement. Detailed diachronic-functional accounts of evidential contrasts found in the Greater Himalayan Region (GHR) reveal that contrasting evidentials are not only defined against each other but also that most of them once had different aspecto-temporal (TA) values which must have aligned when their contrast was conventionalized. Based on these accounts, the present paper sheds light on hitherto unidentified mechanisms of grammatical change. The main insights of the present study were facilitated by ‘functional reconstruction’, which (i) revolves around morphemes which appear to be used in divergent ways within a language and/or across different related languages, (ii) persistently devises hypotheses as to how these functional divergences may have developed, and (iii) retains those hypotheses which most plausibly and economically account for the data. Based on the dense and detailed grammatical literature on the Tibetic language family, the author of this study is able to reconstruct the initial steps by which its evidentiality systems developed: By the time Proto-Tibetan started to be spread across much of Central Asia in the 7th century CE, verbal concatenations with and without a connective -s had become common. As typical for resultative constructions around the globe, Proto-Tibetan *V-s-’dug ‘was there, having undergone V’ (employing the simple past of ’dug ‘stay, be there’) allowed both for a perfect reading (‘the state resulting from V holds at the moment of speech’) and an inferential reading (‘(I infer from its result that) V has taken place’). In Western Tibetic, *V-s-’dug grammaticalized in its perfect meaning as it became contrasted with perfect *V-s-yod ‘is there, having undergone V’ (employing the existential copula yod); that is, *V-s-’dug came to mean that the speaker directly witnessed the profiled result of V, whereas *V-s-yod came to mean that the speaker does not depend on direct evidence of the result, as s/he simply knows that it holds. In Eastern Tibetic, on the other hand, V-s-’dug grammaticalized in its inferential past meaning as it became contrasted with past *V-thal ‘went past V-ing’ (employing the simple past of thal ‘go past’); that is, *V-s-’dug came to mean that the profiled past event was inferred from its result, while *V-thal came to mean that it was directly witnessed. Hence, depending on whether it became contrasted with a perfect or a past construction, resultative V-s-’dug grammaticalized either its direct evidential perfect or its inferential past function. This means that in both cases, evidential readings of constructions with distinct but overlapping TA-values became contrasted, and in order for their contrasting meanings to grammaticalize, the constructions had to agree on their tertium comparationis, which was their shared TA-value. By showing that other types of evidential contrasts in the GHR are also TA-aligned, while no single markers (or privative contrasts) are found to have grammaticalized evidential functions, the present study suggests that, at least in this region of the world, evidential meanings grammaticalize only in equipollent contrasts, which always end up TA-aligned.

Keywords: evidential contrasts, functional-diachronic accounts, grammatical change, himalayan languages, tense/aspect-alignment

Procedia PDF Downloads 94

Authors: Yishu Gong, Liangliang Yang, Jianyu Zhang, Zhengyu Chen, Sihong He, Xusheng Zhang, Wei Zhang

Abstract:

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects millions of people worldwide and is characterized by cognitive decline and behavioral changes. People living with Alzheimer’s disease often find it hard to complete routine tasks. However, there are limited objective assessments that aim to quantify the difficulty of certain tasks for AD patients compared to non-AD people. In this study, we propose to use speech emotion recognition (SER), especially the frustration level, as a potential biomarker for quantifying the difficulty patients experience when describing a picture. We build an SER model using data from the IEMOCAP dataset and apply the model to the DementiaBank data to detect the AD/non-AD group difference and perform longitudinal analysis to track the AD disease progression. Our results show that the frustration level detected from the SER model can possibly be used as a cost-effective tool for objective tracking of AD progression in addition to the Mini-Mental State Examination (MMSE) score.

Keywords: Alzheimer’s disease, speech emotion recognition, longitudinal biomarker, machine learning

Procedia PDF Downloads 77

Authors: J. K. Adedeji, M. O. Oyekanmi

Abstract:

This paper has critically examined the use of Machine Learning procedures in curbing unauthorized access into valuable areas of an organization. The use of passwords, pin codes, user’s identification in recent times has been partially successful in curbing crimes involving identities, hence the need for the design of a system which incorporates biometric characteristics such as DNA and pattern recognition of variations in facial expressions. The facial model used is the OpenCV library which is based on the use of certain physiological features, the Raspberry Pi 3 module is used to compile the OpenCV library, which extracts and stores the detected faces into the datasets directory through the use of camera. The model is trained with 50 epoch run in the database and recognized by the Local Binary Pattern Histogram (LBPH) recognizer contained in the OpenCV. The training algorithm used by the neural network is back propagation coded using python algorithmic language with 200 epoch runs to identify specific resemblance in the exclusive OR (XOR) output neurons. The research however confirmed that physiological parameters are better effective measures to curb crimes relating to identities.

Keywords: biometric characters, facial recognition, neural network, OpenCV

Procedia PDF Downloads 227

Authors: Kothuri Sriraman, Mattupalli Komal Teja

Abstract:

In this paper, a novel algorithm is proposed for the recognition of hulls in a hand written images that might be irregular or digit or character shape. Identification of objects and internal objects is quite difficult to extract, when the structure of the image is having bulk of clusters. The estimation results are easily obtained while going through identifying the sub-regional objects by using the SASK algorithm. Focusing mainly to recognize the number of internal objects exist in a given image, so as it is shadow-free and error-free. The hard clustering and density clustering process of obtained image rough set is used to recognize the differentiated internal objects, if any. In order to find out the internal hull regions it involves three steps pre-processing, Boundary Extraction and finally, apply the Hull Detection system. By detecting the sub-regional hulls it can increase the machine learning capability in detection of characters and it can also be extend in order to get the hull recognition even in irregular shape objects like wise black holes in the space exploration with their intensities. Layered hulls are those having the structured layers inside while it is useful in the Military Services and Traffic to identify the number of vehicles or persons. This proposed SASK algorithm is helpful in making of that kind of identifying the regions and can useful in undergo for the decision process (to clear the traffic, to identify the number of persons in the opponent’s in the war).

Keywords: chain code, Hull regions, Hough transform, Hull recognition, Layered Outline Extraction, SASK algorithm

Procedia PDF Downloads 301

Authors: António J. R. Neves, Rui Garcia, Paulo Dias, Alina Trifan

Abstract:

With the aging of the world population and the continuous growth in technology, service robots are more and more explored nowadays as alternatives to healthcare givers or personal assistants for the elderly or disabled people. Any service robot should be capable of interacting with the human companion, receive commands, navigate through the environment, either known or unknown, and recognize objects. This paper proposes an approach for object recognition based on the use of depth information and color images for a service robot. We present a study on two of the most used methods for object detection, where 3D data is used to detect the position of objects to classify that are found on horizontal surfaces. Since most of the objects of interest accessible for service robots are on these surfaces, the proposed 3D segmentation reduces the processing time and simplifies the scene for object recognition. The first approach for object recognition is based on color histograms, while the second is based on the use of the SIFT and SURF feature descriptors. We present comparative experimental results obtained with a real service robot.

Keywords: object detection, feature, descriptors, SIFT, SURF, depth images, service robots

Procedia PDF Downloads 505

Authors: Vishwanath Pethri Kamath, Jayantha Gowda Sarapanahalli, Vishal Mishra, Siddhesh Balwant Bandgar

Abstract:

Recognition of emotional information is essential in any form of communication. Growing HCI (Human-Computer Interaction) in recent times indicates the importance of understanding of emotions expressed and becomes crucial for improving the system or the interaction itself. In this research work, textual data for emotion recognition is used. The text being the least expressive amongst the multimodal resources poses various challenges such as contextual information and also sequential nature of the language construction. In this research work, the proposal is made for a neural architecture to resolve not less than 8 emotions from textual data sources derived from multiple datasets using google pre-trained word2vec word embeddings and a Multi-head attention-based bidirectional LSTM model with a one-vs-all Multi-Level Classification. The emotions targeted in this research are Anger, Disgust, Fear, Guilt, Joy, Sadness, Shame, and Surprise. Textual data from multiple datasets were used for this research work such as ISEAR, Go Emotions, Affect datasets for creating the emotions’ dataset. Data samples overlap or conflicts were considered with careful preprocessing. Our results show a significant improvement with the modeling architecture and as good as 10 points improvement in recognizing some emotions.

Keywords: text emotion recognition, bidirectional LSTM, multi-head attention, multi-level classification, google word2vec word embeddings

Procedia PDF Downloads 142

Authors: Mohammed S. Al-Rawi, J. Bastos, J. Rodriguez

Abstract:

Object detection and object recognition are essential components of every computer vision system. Despite the high computational complexity and other problems related to numerical stability and accuracy, Zernike moments of 2D images (ZMs) have shown resilience when used in object recognition and have been used in various image analysis applications. In this work, we propose a novel method for computing ZMs via Fast Fourier Transform (FFT). Notably, this is the first algorithm that can generate ZMs up to extremely high orders accurately, e.g., it can be used to generate ZMs for orders up to 1000 or even higher. Furthermore, the proposed method is also simpler and faster than the other methods due to the availability of FFT software and/or hardware. The accuracies and numerical stability of ZMs computed via FFT have been confirmed using the orthogonality property. We also introduce normalizing ZMs with Neumann factor when the image is embedded in a larger grid, and color image reconstruction based on RGB normalization of the reconstructed images. Astonishingly, higher-order image reconstruction experiments show that the proposed methods are superior, both quantitatively and subjectively, compared to the q-recursive method.

Keywords: Chebyshev polynomial, fourier transform, fast algorithms, image recognition, pseudo Zernike moments, Zernike moments

Procedia PDF Downloads 234

Authors: Valentina Zhang

Abstract:

While facial expression is a complex and individualized behavior, all facial emotion recognition (FER) systems known to us rely on a single facial representation and are trained on universal data. We conjecture that: (i) different facial representations can provide different, sometimes complementing views of emotions; (ii) when employed collectively in a discussion group setting, they enable more accurate emotion reading which is highly desirable in autism care and other applications context sensitive to errors. In this paper, we first study FER using pixel-based DL vs semantics-based DL in the context of deepfake videos. Our experiment indicates that while the semantics-trained model performs better with articulated facial feature changes, the pixel-trained model outperforms on subtle or rare facial expressions. Armed with these findings, we have constructed an adaptive FER system learning from both types of models for dyadic or small interacting groups and further leveraging the synthesized group emotions as the ground truth for individualized FER training. Using a collection of group conversation videos, we demonstrate that FER accuracy and personalization can benefit from such an approach.

Keywords: neurodivergence care, facial emotion recognition, deep learning, ground truth for supervised learning

Procedia PDF Downloads 99

Authors: B. Afsharipoor, E. Nazemi

Abstract:

Artificial neural networks (ANNs) are powerful tool for prediction which can be trained based on a set of examples and thus, it would be useful for nonlinear image processing. The present paper reviews several paper regarding applications of ANN in image processing to shed the light on advantage and disadvantage of ANNs in this field. Different steps in the image processing chain including pre-processing, enhancement, segmentation, object recognition, image understanding and optimization by using ANN are summarized. Furthermore, results on using multi artificial neural networks are presented.

Keywords: neural networks, image processing, segmentation, object recognition, image understanding, optimization, MANN

Procedia PDF Downloads 359

Authors: Rubin Dan, Xingcai Wang, Ziyang Chen

Abstract:

A chaotic time series noise reduction method based on the fusion of the local projection method, wavelet transform, and particle swarm algorithm (referred to as the LW-PSO method) is proposed to address the problem of false recognition due to noise in the recognition process of chaotic time series containing noise. The method first uses phase space reconstruction to recover the original dynamical system characteristics and removes the noise subspace by selecting the neighborhood radius; then it uses wavelet transform to remove D1-D3 high-frequency components to maximize the retention of signal information while least-squares optimization is performed by the particle swarm algorithm. The Lorenz system containing 30% Gaussian white noise is simulated and verified, and the phase space, SNR value, RMSE value, and K value of the 0-1 test method before and after noise reduction of the Schreiber method, local projection method, wavelet transform method, and LW-PSO method are compared and analyzed, which proves that the LW-PSO method has a better noise reduction effect compared with the other three common methods. The method is also applied to the classical system to evaluate the noise reduction effect of the four methods and the original system identification effect, which further verifies the superiority of the LW-PSO method. Finally, it is applied to the Chengdu rainfall chaotic sequence for research, and the results prove that the LW-PSO method can effectively reduce the noise and improve the chaos recognition rate.

Keywords: Schreiber noise reduction, wavelet transform, particle swarm optimization, 0-1 test method, chaotic sequence denoising

Procedia PDF Downloads 159

Authors: Hamdi Amroun, Yacine Benziani, Mehdi Ammi

Abstract:

In this paper, we want to determine whether the accurate prediction of nicotine concentration can be obtained by using a network of smart objects and an e-cigarette. The approach consists of, first, the recognition of factors influencing smoking cessation such as physical activity recognition and participant’s behaviors (using both smartphone and smartwatch), then the prediction of the configuration of the e-cigarette (in terms of nicotine concentration, power, and resistance of e-cigarette). The study uses a network of commonly connected objects; a smartwatch, a smartphone, and an e-cigarette transported by the participants during an uncontrolled experiment. The data obtained from sensors carried in the three devices were trained by a Long short-term memory algorithm (LSTM). Results show that our LSTM-based model allows predicting the configuration of the e-cigarette in terms of nicotine concentration, power, and resistance with a root mean square error percentage of 12.9%, 9.15%, and 11.84%, respectively. This study can help to better control consumption of nicotine and offer an intelligent configuration of the e-cigarette to users.

Keywords: Iot, activity recognition, automatic classification, unconstrained environment

Procedia PDF Downloads 195

Authors: Reza Shakoori

Abstract:

This paper presents a structural correction of Arabic and Persian strokes using manipulation of their chain codes in order to improve the rate and performance of Persian and Arabic handwritten word recognition systems. It collects pure and effective features to represent a character with one consolidated feature vector and reduces variations in order to decrease the number of training samples and increase the chance of successful classification. Our results also show that how the proposed approaches can simplify classification and consequently recognition by reducing variations and possible noises on the chain code by keeping orientation of characters and their backbone structures.

Keywords: Arabic, chain code normalization, OCR systems, image processing

Procedia PDF Downloads 367

Authors: Han Xue, Zhang Lanyue

Abstract:

In order to seek more effective feature extraction technology, feature extraction method based on MFCC combined with vector hydrophone is exposed in the paper. The sound pressure signal and particle velocity signal of two kinds of ships are extracted by using MFCC and its evolution form, and the extracted features are fused by using fisher-ratio and correlated distance criterion. The features are then identified by BP neural network. The results showed that MFCC, First-Order Differential MFCC and Second-Order Differential MFCC features can be used as effective features for recognition of underwater targets, and the fusion feature can improve the recognition rate. Moreover, the results also showed that the recognition rate of the particle velocity signal is higher than that of the sound pressure signal, and it reflects the superiority of vector signal processing.

Keywords: vector information, MFCC, differential MFCC, fusion feature, BP neural network

Procedia PDF Downloads 484

Authors: Zainab S. Abdullahi, Zakariyya H. Abdullahi, Sahnun Dahiru

Abstract:

Student attendance in schools is a very important aspect in school management record. In recent years, security systems have become one of the most demanding systems in school. Every institute have its own method of taking attendance, many schools in Nigeria use the old fashion way of taking attendance. That is writing the students name and registration number in a paper and submitting it to the lecturer at the end of the lecture which is time-consuming and insecure, because some students can write for their friends without the lecturer’s knowledge. In this paper, we propose a system that takes attendance using face recognition. There are many automatic methods available for this purpose i.e. biometric attendance, but they all waste time, because the students have to follow a queue to put their thumbs on a scanner which is time-consuming. This attendance is recorded by using a camera attached in front of the class room and capturing the student images, detect the faces in the image and compare the detected faces with database and mark the attendance. The principle component analysis was used to recognize the faces detected with a high accuracy rate. The paper reviews the related work in the field of attendance system, then describe the system architecture, software algorithm and result.

Keywords: attendance system, face detection, face recognition, PCA

Procedia PDF Downloads 323

Authors: Girma Yohannis Bade, Olga Kolesnikova, Grigori Sidorov

Abstract:

Named entity recognition (NER) is a building block for information extraction. Though the information extraction process has been automated using a variety of techniques to find and extract a piece of relevant information from unstructured documents, the discovery of targeted knowledge still poses a number of research difficulties because of the variability and lack of structure in Web data. NER, a subtask of information extraction (IE), came to exist to smooth such difficulty. It deals with finding the proper names (named entities), such as the name of the person, country, location, organization, dates, and event in a document, and categorizing them as predetermined labels, which is an initial step in IE tasks. This survey paper presents the roles and importance of NER to IE from the perspective of different algorithms and application area domains. Thus, this paper well summarizes how researchers implemented NER in particular application areas like finance, medicine, defense, business, food science, archeology, and so on. It also outlines the three types of sequence labeling algorithms for NER such as feature-based, neural network-based, and rule-based. Finally, the state-of-the-art and evaluation metrics of NER were presented.

Keywords: the role of NER, named entity recognition, information extraction, sequence labeling algorithms, named entity application area

Procedia PDF Downloads 41

Authors: Reza Aghayan

Abstract:

Numerically invariant signatures were introduced as a new paradigm of the invariant recognition for visual objects modulo a certain group of transformations. This paper shows that the current formulation suffers from noise and indeterminacy in the resulting joint group-signatures and applies the n-difference technique and the m-mean signature method to minimize their effects. In our experimental results of applying the proposed numerical scheme to generate joint group-invariant signatures, the sensitivity of some parameters such as regularity and mesh resolution used in the algorithm will also be examined. Finally, several interesting observations are made.

Keywords: Euclidean and affine geometry, differential invariant G-signature curves, numerically invariant joint G-signatures, object recognition, noise, indeterminacy

Procedia PDF Downloads 361