Sensor-based activity recognition can recognize simple activities such as walking and running with high accuracy, but it is difficult to recognize complex activities such as nursing care activities and cooking activit...
详细信息
ISBN:
(数字)9789811903618
ISBN:
(纸本)9789811903618;9789811903601
Sensor-based activity recognition can recognize simple activities such as walking and running with high accuracy, but it is difficult to recognize complex activities such as nursing care activities and cooking activities. One solution is to use multiple sensors, which is unrealistic in real life. Recently, learning using privileged information (LUPI) has been proposed, which enables training using additional information only in the training phase. In this paper, we used LUPI for improving the accuracy of complex activity recognition. In short, training is performed with multiple sensors during the training phase, and a single sensor is used during testing. We used four published datasets for evaluating our proposed method. As a result, our proposed method improves by up to 16% in F1-Score to 67% compared to the baseline method when we used random-split cross-validation of each subject.
The proceedings contain 13 papers. The special focus in this conference is on Reproducible Research in patternrecognition. The topics include: Reproducibility Aspects of Crack Detection as a Weakly-Supervised Problem...
ISBN:
(纸本)9783030764227
The proceedings contain 13 papers. The special focus in this conference is on Reproducible Research in patternrecognition. The topics include: Reproducibility Aspects of Crack Detection as a Weakly-Supervised Problem: Towards Achieving Less Annotation-Intensive Crack Detectors;reproducing the Sparse Huffman Address Map Compression for Deep Neural Networks;implementation of Genetic Pseudo Rehearsal;reproducibility: Evaluating the Evaluations;torchdistill: A Modular, Configuration-Driven Framework for Knowledge Distillation;spatio-Temporal Convolutional Autoencoders for Perimeter Intrusion Detection;creating Emotion recognition Algorithms Based on a Convolutional Neural Network for Sentiment Analysis;Tree Defect Segmentation Using Geometric Features and CNN;pith Estimation on Tree Log End Images;structure, Concept and Result Reproducibility of the Benchmark on Vesselness Filters;a Heuristic-Based Decision Tree for Connected Components Labeling of 3D Volumes: Implementation and Reproducibility Notes.
The characteristics of various traffic modes are analyzed, the fuzzy neural network is used to identify the traffic modes, the eigenvalues of the identified traffic modes are determined, and the correctness of the net...
The characteristics of various traffic modes are analyzed, the fuzzy neural network is used to identify the traffic modes, the eigenvalues of the identified traffic modes are determined, and the correctness of the network is tested by traffic flow. The manual approaches significantly capture the spatial and temporal relationships as well as enhance the model complexity through examining both connected and unconnected roads. In feature extraction process, initially the input image is provided to the computation gradients process and processed output provided to the collect HOG over image and it provides the HOG feature. The test results show that the application of fuzzy neural network to traffic patternrecognition can accurately reflect the situation of traffic flow. Taking the results of traffic patternrecognition as the basis of road patency can guide the group control system to adopt corresponding strategies according to different traffic conditions.
The proceedings contain 92 papers. The topics discussed include: transformer health index by prediction artificial neural networks diagnostic techniques;energy efficient ant colony system for packet routing in wireles...
The proceedings contain 92 papers. The topics discussed include: transformer health index by prediction artificial neural networks diagnostic techniques;energy efficient ant colony system for packet routing in wireless sensor network;evaluation of two-part rain attenuation model at Ku-band for tropical and equatorial regions;performance of restricted fault and bias differential protection against earth fault on a transformer;design and implementation of an automatic speed control system of vehicles for avoiding road accidents in Bangladesh;real time retinal optic disc segmentation via guided filter and discrete wavelet transform;air plasma sterilizer using a parallel dielectric barrier discharge;design and characterization of screen-printed piezoresistive cantilever for gas sensor application;effect of build parameters on process energy consumption and material usage in fused deposition modelling method;and channels selection for patternrecognition of five fingers motor imagery electroencephalography signals.
This work addresses the development of a hybrid method for feature selection and a strategy to classify quite large datasets of handwritten characters. The divide and conquer paradigm, generally, is used to divide a b...
详细信息
ISBN:
(纸本)9781665455176
This work addresses the development of a hybrid method for feature selection and a strategy to classify quite large datasets of handwritten characters. The divide and conquer paradigm, generally, is used to divide a big problem into minor problems. This research applied this concept to recognize handwritten uppercase letters and numbers. As a result, a big problem is split into two nodes or subproblems, one for numbers and one for letters. Then, letters are divided into two nodes representing the straight and curved ones. The division can be called the binary decision tree and allows to obtain a subset with the minimal features of each node called reduct. Here, an improvement of reducts is proposed using the ant colony algorithm as the embedded method. The application of these methods had the following result and conclusions. For each node, subsets of fewer features were obtained with high performance in the classification, considering the morphology of each letter. It is crucial to highlight that the distribution of the samples affects the performance of the classifier and the strategy improves the performance of the reduct.
The proceedings contain 105 papers. The topics discussed include: the effect of hg content in water bodies for precision medicine of Hg disease;the spatial migration pattern of Hg content for the prediction of disease...
ISBN:
(纸本)9781450398442
The proceedings contain 105 papers. The topics discussed include: the effect of hg content in water bodies for precision medicine of Hg disease;the spatial migration pattern of Hg content for the prediction of disease during a long time;Cryptophycin’s clinical use in xenograft pancreatic tumor cell treatments based on computer science;standardization of clinical terminology based on hybrid recall and Ernie;interaction process between changes in terms of time and space of pH value and the growth of marine organisms in the Jiaozhou Bay;clinical research progress of acupuncture in the treatment of sciatica;research on identification of abnormal foot arch in school-age children based on foot pressure analysis;intelligent diagnosis of vascular anomalies with deep learning;the value of brain-computer interface in stroke upper rehabilitation;and research on abnormal behavior recognition of the elderly based on spatial-temporal feature fusion.
In this paper, we propose an approach that improves segmentation networks with automatic augmentation networks for dental mesh data. Since conventional data augmentation is to augment all samples uniformly with predef...
详细信息
Understanding document images uploaded on social media is challenging because of multiple types like handwritten, printed and scene text images. This study presents a new model called Deep Fuzzy based MSER for classif...
详细信息
We proposes a new kind of contrast learning method based on Temporal-Correlation characters of electroencephalogram (EEG) used in sleep staging. This method, called simple framework of temporal-correlation representat...
详细信息
Finding stolen cars is becoming increasingly important in many urban regions. An automated system for scanning license plates can recognize vehicle numbers without the need for human interaction. This work proposes a ...
Finding stolen cars is becoming increasingly important in many urban regions. An automated system for scanning license plates can recognize vehicle numbers without the need for human interaction. This work proposes a vehicle theft detection system based on neural patternrecognition, gaussian filter, and equilibrium optimization. The proposed architecture has significant speedup and higher accuracy rates. The proposed number plate recognition method has a maximum accuracy rate of 94% and an average reduction in the processing time of 32%. patternrecognition is the process of detecting regularities and patterns in data using machine learning. These analogies may now be uncovered via statistical analysis, historical data, or machine-generated knowledge.
暂无评论