CNN-based multi-class multi-label classification of sound scenes in the context of wind turbine sound emission measurements

2021 ◽  
Vol 263 (4) ◽  
pp. 2687-2698
Author(s):  
Nils Poschadel ◽  
Christian Gill ◽  
Stephan Preihs ◽  
Jürgen Peissig
Keyword(s):  
Wind Turbine ◽  
Wind Turbines ◽  
Classification Model ◽  
Automated Classification ◽  
Sound Emission ◽  
Audio Signals ◽  
Reliable Classification ◽  
Acoustic Conditions ◽  
Manual Classification

Within the scope of the interdisciplinary project WEA-Akzeptanz, measurements of the sound emission of wind turbines were carried out at the Leibniz University Hannover. Due to the environment there are interfering components (e. g. traffic, birdsong, wind, rain, ...) in the recorded signals. Depending on the subsequent signal processing and analysis, it may be necessary to identify sections with the raw sound of a wind turbine, recordings with the purest possible background noise or even a specific combination of interfering noises. Due to the amount of data, a manual classification of the audio signals is usually not feasible and an automated classification becomes necessary. In this paper, we extend our previously proposed multi-class single-label classification model to a multi-class multi-label model, which reflects the real-world acoustic conditions around wind turbines more accurately and allows for finer-grained evaluations. We first provide a short overview of the data acquisition and the dataset. We then briefly summarize our previous approach, extend it to a multi-class multi-label formulation, and analyze the trained convolutional neural network regarding different metrics. All in all, the model delivers very reliable classification results with an overall example-based F1-score of about 80 % for a multi-label classification of 12 classes.

scholarly journals Automatic classification of older electronic texts into the Universal Decimal Classification–UDC

2020 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Matjaž Kragelj ◽  
Mirjana Kljajić Borštnar
Keyword(s):  
Machine Learning ◽  
Digital Libraries ◽  
Classification Model ◽  
Research Approach ◽  
Automated Classification ◽  
Content Type ◽  
Machine Learning Classification ◽  
Universal Decimal Classification ◽  
Decimal Classification

PurposeThe purpose of this study is to develop a model for automated classification of old digitised texts to the Universal Decimal Classification (UDC), using machine-learning methods.Design/methodology/approachThe general research approach is inherent to design science research, in which the problem of UDC assignment of the old, digitised texts is addressed by developing a machine-learning classification model. A corpus of 70,000 scholarly texts, fully bibliographically processed by librarians, was used to train and test the model, which was used for classification of old texts on a corpus of 200,000 items. Human experts evaluated the performance of the model.FindingsResults suggest that machine-learning models can correctly assign the UDC at some level for almost any scholarly text. Furthermore, the model can be recommended for the UDC assignment of older texts. Ten librarians corroborated this on 150 randomly selected texts.Research limitations/implicationsThe main limitations of this study were unavailability of labelled older texts and the limited availability of librarians.Practical implicationsThe classification model can provide a recommendation to the librarians during their classification work; furthermore, it can be implemented as an add-on to full-text search in the library databases.Social implicationsThe proposed methodology supports librarians by recommending UDC classifiers, thus saving time in their daily work. By automatically classifying older texts, digital libraries can provide a better user experience by enabling structured searches. These contribute to making knowledge more widely available and useable.Originality/valueThese findings contribute to the field of automated classification of bibliographical information with the usage of full texts, especially in cases in which the texts are old, unstructured and in which archaic language and vocabulary are used.

scholarly journals Reliable Classification of FAQs with Spelling Errors Using an Encoder-Decoder Neural Network in Korean

2019 ◽  
Vol 9 (22) ◽  
pp. 4758 ◽  
Author(s):  
Youngjin Jang ◽  
Harksoo Kim
Keyword(s):  
Neural Network ◽  
Domain Knowledge ◽  
Classification Model ◽  
Word Embeddings ◽  
Spelling Errors ◽  
Proposed Model ◽  
Sentence Classification ◽  
Improved Performance ◽  
Reliable Classification

To resolve lexical disagreement problems between queries and frequently asked questions (FAQs), we propose a reliable sentence classification model based on an encoder-decoder neural network. The proposed model uses three types of word embeddings; fixed word embeddings for representing domain-independent meanings of words, fined-tuned word embeddings for representing domain-specific meanings of words, and character-level word embeddings for bridging lexical gaps caused by spelling errors. It also uses class embeddings to represent domain knowledge associated with each category. In the experiments with an FAQ dataset about online banking, the proposed embedding methods contributed to an improved performance of the sentence classification. In addition, the proposed model showed better performance (with an accuracy of 0.810 in the classification of 411 categories) than that of the comparison model.

scholarly journals Automated classification of stages of anaesthesia by populations of evolutionary optimized fuzzy rules

2015 ◽  
Vol 1 (1) ◽  
pp. 77-79
Author(s):  
C. Walther ◽  
A. Wenzel ◽  
M. Schneider ◽  
M. Trommer ◽  
K.-P. Sturm ◽  
...  
Keyword(s):  
Evolutionary Algorithms ◽  
Embedded System ◽  
Vital Signs ◽  
Fuzzy Rule ◽  
Fuzzy Rules ◽  
Classification Model ◽  
Multiobjective Evolutionary Algorithms ◽  
Automated Classification ◽  
Rule Sets

AbstractThe detection of stages of anaesthesia is mainly performed on evaluating the vital signs of the patient. In addition the frontal one-channel electroencephalogram can be evaluated to increase the correct detection of stages of anaesthesia. As a classification model fuzzy rules are used. These rules are able to classify the stages of anaesthesia automatically and were optimized by multiobjective evolutionary algorithms. As a result the performance of the generated population of fuzzy rule sets is presented. A concept of the construction of an autonomic embedded system is introduced. This system should use the generated rules to classify the stages of anaesthesia using the frontal one-channel electroencephalogram only.

scholarly journals Automatic classification of ICA components from infant EEG using MARA

2021 ◽  
Author(s):  
I. Marriot Haresign ◽  
E. Phillips ◽  
M. Whitehorn ◽  
V. Noreika ◽  
E.J.H. Jones ◽  
...  
Keyword(s):  
Visual Processing ◽  
Free Play ◽  
Automatic Identification ◽  
Automated Classification ◽  
Specific System ◽  
Flexible Tool ◽  
Eeg Data ◽  
Ocular Artifact ◽  
Manual Classification

AbstractAutomated systems for identifying and removing non-neural ICA components are growing in popularity among adult EEG researchers. Infant EEG data differs in many ways from adult EEG data, but there exists almost no specific system for automated classification of source components from paediatric populations. Here, we adapt one of the most popular systems for adult ICA component classification for use with infant EEG data. Our adapted classifier significantly outperformed the original adult classifier on samples of naturalistic free play EEG data recorded from 10 to 12-month-old infants, achieving agreement rates with the manual classification of over 75% across two validation studies (n=44, n=25). Additionally, we examined both classifiers ability to remove stereotyped ocular artifact from a basic visual processing ERP dataset, compared to manual ICA data cleaning. Here the new classifier performed on level with expert manual cleaning and was again significantly better than the adult classifier at removing artifact whilst retaining a greater amount of genuine neural signal, operationalised through comparing ERP activations in time and space. Our new system (iMARA) offers developmental EEG researchers a flexible tool for automatic identification and removal of artifactual ICA components.

scholarly journals Big transfer learning for automated skin cancer classification

2021 ◽  
Vol 23 (3) ◽  
pp. 1611
Author(s):  
Zinah Mohsin Arkah ◽  
Dalya S. Al-Dulaimi ◽  
Ahlam R. Khekan
Keyword(s):  
Deep Learning ◽  
Skin Cancer ◽  
Transfer Learning ◽  
Image Features ◽  
Cancer Classification ◽  
Automated Classification ◽  
Novel Approach ◽  
Learned Features ◽  
Manual Classification

<p>Skin cancer is an example of the most dangerous disease. Early diagnosis of skin cancer can save many people’s lives. Manual classification methods are time-consuming and costly. Deep learning has been proposed for the automated classification of skin cancer. Although deep learning showed impressive performance in several medical imaging tasks, it requires a big number of images to achieve a good performance. The skin cancer classification task suffers from providing deep learning with sufficient data due to the expensive annotation process and required experts. One of the most used solutions is transfer learning of pre-trained models of the ImageNet dataset. However, the learned features of pre-trained models are different from skin cancer image features. To end this, we introduce a novel approach of transfer learning by training the pre-trained models of the ImageNet (VGG, GoogleNet, and ResNet50) on a large number of unlabelled skin cancer images, first. We then train them on a small number of labeled skin images. Our experimental results proved that the proposed method is efficient by achieving an accuracy of 84% with ResNet50 when directly trained with a small number of labeled skin and 93.7% when trained with the proposed approach.</p>

EXPERIMENTAL AND ANALYTICAL STUDIES OF VERTICAL AXIS WIND TURBINES

2018 ◽  
pp. 51-58
Author(s):  
B. P. Khozyainov
Keyword(s):  
Wind Turbine ◽  
Flow Velocity ◽  
Wind Power ◽  
Wind Turbines ◽  
Power Efficiency ◽  
Air Flow ◽  
Geometrical Parameters ◽  
Wind Speeds ◽  
Analytical Studies ◽  
Air Flow Velocity

The article carries out the experimental and analytical studies of three-blade wind power installation and gives the technique for measurements of angular rate of wind turbine rotation depending on the wind speeds, the rotating moment and its power. We have made the comparison of the calculation results according to the formulas offered with the indicators of the wind turbine tests executed in natural conditions. The tests were carried out at wind speeds from 0.709 m/s to 6.427 m/s. The wind power efficiency (WPE) for ideal traditional installation is known to be 0.45. According to the analytical calculations, wind power efficiency of the wind turbine with 3-bladed and 6 wind guide screens at wind speedsfrom 0.709 to 6.427 is equal to 0.317, and in the range of speed from 0.709 to 4.5 m/s – 0.351, but the experimental coefficient is much higher. The analysis of WPE variations shows that the work with the wind guide screens at insignificant average air flow velocity during the set period of time appears to be more effective, than the work without them. If the air flow velocity increases, the wind power efficiency gradually decreases. Such a good fit between experimental data and analytical calculations is confirmed by comparison of F-test design criterion with its tabular values. In the design of wind turbines, it allows determining the wind turbine power, setting the geometrical parameters and mass of all details for their efficient performance.

SHAPING OF AN AUTONOMOUS POWER SYSTEM FOR GUARANTEED POWER SUPPLY WITH THE PREDOMINANCE WIND ENERGY

2018 ◽  
pp. 28-50
Author(s):  
S. G. Ignatiev ◽  
S. V. Kiseleva
Keyword(s):  
Energy Storage ◽  
Wind Speed ◽  
Wind Energy ◽  
Wind Turbine ◽  
Wind Turbines ◽  
Energy Demand ◽  
Diesel Generator ◽  
Average Wind Speed ◽  
Wind Speeds ◽  
Average Wind

Optimization of the autonomous wind-diesel plants composition and of their power for guaranteed energy supply, despite the long history of research, the diversity of approaches and methods, is an urgent problem. In this paper, a detailed analysis of the wind energy characteristics is proposed to shape an autonomous power system for a guaranteed power supply with predominance wind energy. The analysis was carried out on the basis of wind speed measurements in the south of the European part of Russia during 8 months at different heights with a discreteness of 10 minutes. As a result, we have obtained a sequence of average daily wind speeds and the sequences constructed by arbitrary variations in the distribution of average daily wind speeds in this interval. These sequences have been used to calculate energy balances in systems (wind turbines + diesel generator + consumer with constant and limited daily energy demand) and (wind turbines + diesel generator + consumer with constant and limited daily energy demand + energy storage). In order to maximize the use of wind energy, the wind turbine integrally for the period in question is assumed to produce the required amount of energy. For the generality of consideration, we have introduced the relative values of the required energy, relative energy produced by the wind turbine and the diesel generator and relative storage capacity by normalizing them to the swept area of the wind wheel. The paper shows the effect of the average wind speed over the period on the energy characteristics of the system (wind turbine + diesel generator + consumer). It was found that the wind turbine energy produced, wind turbine energy used by the consumer, fuel consumption, and fuel economy depend (close to cubic dependence) upon the specified average wind speed. It was found that, for the same system with a limited amount of required energy and high average wind speed over the period, the wind turbines with lower generator power and smaller wind wheel radius use wind energy more efficiently than the wind turbines with higher generator power and larger wind wheel radius at less average wind speed. For the system (wind turbine + diesel generator + energy storage + consumer) with increasing average speed for a given amount of energy required, which in general is covered by the energy production of wind turbines for the period, the maximum size capacity of the storage device decreases. With decreasing the energy storage capacity, the influence of the random nature of the change in wind speed decreases, and at some values of the relative capacity, it can be neglected.

Connectionist Temporal Classification Model for Dynamic Hand Gesture Recognition using RGB and Optical flow Data

2020 ◽  
Vol 17 (4) ◽  
pp. 497-506
Author(s):  
Sunil Patel ◽  
Ramji Makwana
Keyword(s):  
Neural Network ◽  
Optical Flow ◽  
Gesture Recognition ◽  
Hand Gesture Recognition ◽  
Classification Model ◽  
Hand Gesture ◽  
Flow Data ◽  
Dynamic Hand Gesture Recognition ◽  
Connectionist Temporal Classification

Automatic classification of dynamic hand gesture is challenging due to the large diversity in a different class of gesture, Low resolution, and it is performed by finger. Due to a number of challenges many researchers focus on this area. Recently deep neural network can be used for implicit feature extraction and Soft Max layer is used for classification. In this paper, we propose a method based on a two-dimensional convolutional neural network that performs detection and classification of hand gesture simultaneously from multimodal Red, Green, Blue, Depth (RGBD) and Optical flow Data and passes this feature to Long-Short Term Memory (LSTM) recurrent network for frame-to-frame probability generation with Connectionist Temporal Classification (CTC) network for loss calculation. We have calculated an optical flow from Red, Green, Blue (RGB) data for getting proper motion information present in the video. CTC model is used to efficiently evaluate all possible alignment of hand gesture via dynamic programming and check consistency via frame-to-frame for the visual similarity of hand gesture in the unsegmented input stream. CTC network finds the most probable sequence of a frame for a class of gesture. The frame with the highest probability value is selected from the CTC network by max decoding. This entire CTC network is trained end-to-end with calculating CTC loss for recognition of the gesture. We have used challenging Vision for Intelligent Vehicles and Applications (VIVA) dataset for dynamic hand gesture recognition captured with RGB and Depth data. On this VIVA dataset, our proposed hand gesture recognition technique outperforms competing state-of-the-art algorithms and gets an accuracy of 86%

Sign in / Sign up

Export Citation Format

Share Document