检索结果-内蒙古大学图书馆

Study on recognition and classification of English accents using deep learning algorithms

JOURNAL OF INTELLIGENT SYSTEMS 2023年第1期32卷

作者： Ke, Wenjuan Hubei Univ Sci & Technol Sch Foreign Languages Bldg 31Tianjieshuhua City Xianning 437100 Hubei Peoples R China

The recognition and classification of English accents have high practical value in areas such as security management and information retrieval. This study introduced two English accent features, filter bank (FBank) and Mel-frequency cepstral coefficient (MFCC), based on deep learning techniques. It then combined convolutional neural network (CNN), gated recurrent unit, and an attention mechanism to design a 1D CNN-BiGRU-Attention model for English accent recognition and classification. Experimental tests were conducted on the VoxForge dataset. The results showed that compared to MFCC, FBank performed better in English accent recognition and classification, and 70FBank achieved the highest F1 value. Among the recurrent neural network, long short-term memory, and other models, the BiGRU model had the best performance. The average F1 value of the 1D CNN-BiGRU-attention model was the highest, reaching 85.52%, and all the F1 values were above 80% for different accents, indicating that the addition of the attention mechanism effectively improved the model's recognition and classification effectiveness. The results prove the reliability of the method proposed in this article for English accent recognition and classification, making it suitable for practical application and promotion.

关键词： deep learning English accent recognition and classification FBank

来源：评论

学校读者我要写书评

暂无评论

DeepSniffer: A meta-learning-based chemiresistive odor sensor for recognition and classification of aroma oils

引用

SENSORS AND ACTUATORS B-CHEMICAL 2022年第0期351卷 130960-130960页

作者： Liu, Chuanjun Miyauchi, Hitoshi Hayashi, Kenshi USE Co Ltd Res Lab Shibuya Ku 22-10 Ebisu 4 Chome Tokyo 1500012 Japan Kyushu Univ Grad Sch Informat Sci & Elect Engn Nishi Ku 744 Motooka Fukuoka 8190395 Japan

A meta-learning algorithm, conventionally used for visual recognition, was applied to the recognition and classification of aroma oils. A printable chemiresistive sensor array was fabricated, based on composites of carbon black with various active materials. Standard aromatherapy kits with 30 types of essential oils were used as targets in an odor sensing experiment. Benefiting from the pattern recognition ability of the fabricated sensor array, a high-quality dataset was obtained with 30 aroma oil classes, in which each class had nine replicate samples. A deep metric learning model, based on a Siamese neural network and a multilayer perceptron, was used to perform the N-way k-shot meta-learning. A test accuracy of over 98.7% was obtained for 31-way 9-shot learning, on discriminating whether the input pair samples were taken from similar or dissimilar classes. The model was effective in extracting meta-features of the aroma oils;this was proved by the improved clustering effect of samples in the spaces of principal components analysis and t-distributed stochastic neighbor embedding. The 30 aroma oils were divided into two datasets according to 6-fold cross-validation: 25 aroma oil classes (plus one blank class) as seen classes for constructing 26-way 9-shot learning models and the remaining five aroma oils as unseen classes for prediction. Average accuracies of 93.5% and 93.9% were achieved for recognition of the unseen aroma oils from the seen classes and classification of the unseen aroma oils themselves, respectively, demonstrating the effectiveness of the developed sensor and model for odor recognition and classification.

关键词： Chemiresistive sensor array Meta learning Deep metric learning Siamese network Aroma oil recognition and classification

来源：评论

学校读者我要写书评

暂无评论

recognition and classification of Wire Bonding Joint via Image Feature and SVM Model

引用

IEEE TRANSACTIONS ON COMPONENTS PACKAGING AND MANUFACTURING TECHNOLOGY 2019年第5期9卷 998-1006页

作者： Long, Zhili Zhou, Xing Zhang, Xiaobing Wang, Rui Wu, Xiaojun Harbin Inst Technol Shenzhen Shenzhen 518055 Peoples R China Guangdong Univ Technol Minist Educ Key Lab Precis Microelect Mfg Technol & Equipment Guangzhou 510006 Guangdong Peoples R China

recognition and classification for wire bonding joint are important to quality assurance in semiconductor device manufacturing. In this paper, a precision recognition and classification system for bonding joint of ultrasonic heavy aluminum wire based on image feature and support vector machine (SVM) is presented. This system consists of feature extraction from images and classification model. In feature extraction, image processing algorithms including Canny edge extraction, histogram equalization, and image morphology closed operation are utilized to extract and locate a joint contour in a complicated background image. In the classification model, the principal component analysis (PCA) is employed to visualize, reconstruct, and reduce the images data dimension for less computation time. The SVM-based model is chosen as the classifier to identify and recognize joint types. The Gauss-radial basis function (RBF) kernel function is adopted in SVM, and its optimal parameters are determined by cross-validation. In the experiment, 588 bonding images are used to implement in this recognition and classification system. The results prove that the classification accuracy for wire bonding joint based on image feature, PCA, and SVMcan achieve to 97.3%, and the computation time can be reduced significantly.

关键词： Principal component analysis (PCA) recognition and classification support vector machine (SVM) model wire bonding joint

来源：评论

学校读者我要写书评

暂无评论

Deep Learning based Framework for Underwater Acoustic Signal recognition and classification 18

Deep Learning based Framework for Underwater Acoustic Signal...

引用

2nd International Conference on Computer Science and Artificial Intelligence (CSAI) / 10th International Conference on Information and Multimedia Technology (ICIMT)

作者： Wu, Hao Song, Qingzeng Jin, Guanghao Tianjin Polytech Univ Sch Comp Sci & Software Engn Tianjin Peoples R China

ISBN: (纸本)9781450366069

The recognition and classification are important to the research contents like underwater acoustic signal processing. Generally, the state-of-the-art signal recognition systems depend on feature extraction that is based on the knowledge or experience of experts, so that, it can efficiently represent target signatures. In contrast, this paper introduces a new framework based on deep learning methods to preprocess signals and extract features for the recognition and classification of underwater acoustic signals, which present the targets of ships and torpedoes. In our framework, signals are firstly transformed to spectral images of LOFAR, so that, those images can be extracted and classified by convolutional neural networks (CNN). The experimental results show that our deep learning based framework can obtain high classification accuracy in underwater acoustic signals case with the transformation to LOFAR spectrum. The accuracy of our best version reaches 97.22%, higher than those that use other networks, and achieved the expected objectives for real applications.

关键词： Underwater Acoustic Signal LOFAR Spectrum Deep Learning recognition and classification

来源：评论

学校读者我要写书评

暂无评论

A New recognition and classification Algorithm of Underwater Acoustic Signals Based on Multi-Domain Features Combination

A New Recognition and Classification Algorithm of Underwater...

引用

IEEE/OES China Ocean Acoustics (COA)

作者： Shi Huan Xiong Jinyu Zhou Chenyang Yang Su Natl Key Lab Sci & Technol Blind Signal Proc Chengdu 610041 Sichuan Peoples R China

ISBN: (纸本)9781467399784

A new algorithm is proposed in this paper to recognize and classify underwater acoustic signals including underwater acoustic communication signals, active sonar signals, ship-radiating noise and burst ocean ambient noise. The paper firstly analyzed the features of underwater acoustic signals in time domain, frequency domain and cyclostationarity domain. At the same time, the way of extracting features about these signals is researched. This algorithm combines the features above and chooses a decision tree classifier to realize the recognition and classification of different underwater acoustic signals. Finally, the effectiveness in different multipath channels is evaluated by simulation. The results have shown that this algorithm is capable of recognizing and classifying different underwater acoustic signals with better performance than algorithms using features in only one domain;and, its performance is robust in a multipath fading environment.

关键词： underwater acoustic signals recognition and classification feature extraction multi-domain features

来源：评论

学校读者我要写书评

暂无评论

Tennis Posture classification and recognition Based on an Improved KNN 12

Tennis Posture Classification and Recognition Based on an Im...

引用

12th International Conference on Graphics and Image Processing (ICGIP)

作者： Wang, Xinran Huang, Yaxin Zhong, Jianpeng Zhu, Yongqi Tang, Qing Wang, Meili Li, Shuqin NorthWest A&F Univ Coll Informat Engn Xianyang Peoples R China Minist Agr & Rural Affairs Key Lab Agr Internet Things Yangling 712100 Shaanxi Peoples R China Shaanxi Key Lab Agr Informat Percept & Intelligen Yangling 712100 Shaanxi Peoples R China

ISBN: (纸本)9781510642782;9781510642775

Motion recognition is widely used in somatosensory games, rehabilitation training and robot motion learning. Tennis training can identify and classify the captured actions, and improve the performance of computer-aided tennis teaching timely and accurately. Traditional image based or video based human posture capture recognition are easily affected by complex background environments, different lighting conditions and other factors in practical applications. In this paper, Microsoft Kinect is used as the sensor device to capture the data information of the tennis player. Firstly, Kinect's depth sensing technology is used to obtain human skeleton data. Secondly, in order to improve the efficiency of classification, this paper reduces the dimension of data by extracting the feature value of human skeletons. Thirdly, a kind of KNN algorithm which defines the dimension weight is proposed to implement the movement classification, compared with the 94% accuracy of the algorithm, the accuracy of the KNN algorithm is 92.4%, the accuracy of the decision tree algorithm is 92.81%, and the accuracy of the CNN algorithm The accuracy is 89.97%. The evaluation method of tennis action is defined to provide guidance for users. By comparing the difference between the user's postures and the standard postures in the joint positions and angles which is prone to get error, this method can correct the user's postures and build up the function of movements guidance. From the teaching effect of tennis aficionados and general tennis players, this method is more practical and targeted than the traditional tennis graphics and video teaching.

关键词： Kinect KNN motion capture recognition and classification

来源：评论

学校读者我要写书评

暂无评论

Simultaneous carbon fiber layer thickness and direction measurement and identification using a novel eddy current sensor and simplified multi-scale 1D-ResNet network

引用

MEASUREMENT 2025年第PartA期242卷

作者： She, Saibo Zheng, Xinnan Zou, Xun Yu, Kuohai Shen, Jialong Wu, Fanfu Yin, Wuliang Guilin Univ Technol Collaborat Innovat Ctr Explorat Nonferrous Met Dep Guilin 541004 Peoples R China Univ Manchester Sch Elect & Elect Engn Oxford Rd Manchester M13 9PL England Univ Warwick WMG Coventry CV4 7AL England

A novel eddy current sensor, featuring a triangular excitation coil and three rectangular receiving coils, is proposed for simultaneously detecting carbon fiber layer thickness and directions in the manufacturing process of carbon fiber reinforced polymer (CFRP) materials. It can disclose various carbon fiber layers and directions based on the signals obtained from the three receiving coils. However, classification of carbon fiber layers and directions proves challenging due to the complexity of signal features resulting from different combinations of carbon fiber layers and directions. To address this issue, we propose the use of a simplified multi-scale 1DResNet network. Comparative analysis reveals that our proposed network achieves higher train and validation accuracy compared to single-scale 1D-ResNet networks and different layers 1D-ResNet networks. Furthermore, the proposed simplified network not only achieves high classification accuracy with 100% and 99.7% for carbon fiber layers and directions, respectively, but also requires less training time compared to multi-scale 1DResNet networks. The novel eddy current sensor proposed in this study enables accurate recognition of carbon fiber layers and their directional orientation through the implementation of a simplified multi-scale 1D-ResNet network in the manufacturing process of CFRP materials. This innovation significantly enhances the quality of CFRP materials during manufacturing, ensuring their strength aligns with predetermined orientations.

关键词： Eddy current testing Carbon fiber thickness and direction Sensor design recognition and classification Deep learning

来源：评论

学校读者我要写书评

暂无评论

TARGET classification WITH OVERLAPPING WINDOW INTEGRATION USING WIGNER VILLE TRANSFORM 31

TARGET CLASSIFICATION WITH OVERLAPPING WINDOW INTEGRATION US...

引用

31st IEEE Conference on Signal Processing and Communications Applications (SIU)

作者： Aksuren, Ibrahim Gokhan Ozer, Erhan Hocaoglu, Ali Koksal Gebze Tekn Univ Elekt Muhendisli Kocaeli Turkiye

ISBN: (纸本)9798350343557

In this study, automatic target classification was performed using underwater acoustic signals. Targets of underwater and surface ships were detected from long distances and classified using a signal integration algorithm with Wigner-Ville distribution for analyzing their noise. The proposed method exhibits an innovative approach by using Wigner-Ville distribution for signal integration and providing early detection and diagnosis compared to existing methods. Additionally, the Wigner-Ville distribution provides better discriminative features by separating frequency lines of targets from background noise with signal integration, which is observed as a new approach. With this proposed approach, it has been demonstrated that target classification performance improves.

关键词： Underwater Acoustics Acoustic Signal Processing Underwater Acoustic Target recognition and classification Wigner Ville Distrubution Signal Integration

来源：评论

学校读者我要写书评

暂无评论

Accurate classification of power quality disturbance based on 3D visualized spiral curve and hybrid ER-MVCNN model

引用

MEASUREMENT 2024年 231卷

作者： Song, Juncai Wu, Jing Wang, Xiaoqing Duan, Zhangling Wang, Xiaoxian Lu, Siliang Anhui Univ Sch Internet Hefei 230039 Anhui Peoples R China Anhui Univ Sch Elect & Informat Engn Hefei 230601 Anhui Peoples R China Anhui Univ Sch Elect Engn & Automat Hefei 230601 Anhui Peoples R China

This work proposes a new method for accurate power quality disturbance (PQD) classification. Firstly, 3D-visualized spiral curve (3D-VSC) is innovatively proposed to convert 1D PQD signal into 3D spiral curve. It can display time-frequency information in multi-dimensional perspectives and the disturbance characteristics more intuitively in image visual sense, which can realize PQD features enhancement. Secondly, efficient channel attention-ResNet multi-view convolution neural network (ER-MVCNN) hybrid model is creatively proposed by introducing ECA-ResNet as feature extraction module. It can solve limitations of multi-view convolution neural network (MVCNN) in single-view feature extraction and defects of multi-view feature fusion in the weighting problem, which can effectively achieve classification accuracy to 99.68 %. Comparison experiments with MVCNN, AlexNet-MVCNN, VGGNet-MVCNN, ResNet18-MVCNN, and SENet-ResNet-MVCNN show that ERMVCNN has higher accuracy by 12.4 %, 6.79 %, 6.14 %, 3.86 %, and 0.88 % respectively. Finally, through establishing PQD experiment platform and IEEE PES datasets to verify high accuracy and robustness of proposed method.

关键词： Power quality disturbance 3D-visualized spiral curve Multi-view convolutional neural network (MVCNN) Efficient channel attention-ResNet multi-view convolution neural network (ER-MVCNN) recognition and classification

来源：评论

学校读者我要写书评

暂无评论

AUTOMATIC TARGET classification USING UNDERWATER ACOUSTIC SIGNALS 30

AUTOMATIC TARGET CLASSIFICATION USING UNDERWATER ACOUSTIC SI...

引用

30th IEEE Signal Processing and Communications Applications Conference (SIU)

作者： Aksuren, Ibrahim Gokhan Hocaoglu, Ali Koksal Gebze Tekn Univ Elekt Muhendisligi Kocaeli Turkey

ISBN: (纸本)9781665450928

In this study, automatic target classification was performed using underwater acoustic signals. The targets were classified using the Low Frequency Analysis and Recording algorithm, which is used for detecting underwater and surface ships from long distances and for target recognition by analyzing their noise. While the proposed method uses 6 different frequency ranges separately for target classification, instead of a single frequency range, comparing to existing methods, it presents an innovative approach by presenting a LOFAR image to the SONAR operator in 7 different frequency ranges. Additionally, LOFAR data is visualized by the t-SNE algorithm and it is observed that grouping the ships based on noise characteristics forms a more compact clusters compared to groupings based on the sizes of the ships. We show that the proposed approach improves the performance of target classification.

关键词： Underwater Acoustics Acoustic Signal Processing Underwater Acoustic Target recognition and classification LOFAR Analysis

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：