Pub Date : 2023-03-01DOI: 10.1109/ICNLP58431.2023.00089
Juanyi Zheng, Yuanyuan Lv, Jinyu Mu, Lirong Xing, Pei Jie
Since the conventional Compressed Sensing (CS) algorithm in millimeter wave Massive Multi-Input Multi-Output (MIMO) system has the problem of low channel estimation accuracy, a deep learning based beamspace channel estimation method, Deep Iterative Shrinkage-Thresholding Algorithm (DISTA), is proposed. First, due to the sparsity of the beamspace channel, the beamspace channel estimation problem can be transformed into a sparse signal recovery problem; second, based on the iterative shrinkage threshold algorithm (ISTA), the channel state information (CSI) is sparsified using nonlinear transformation functions to replace the traditional manual transformation; finally, the iterative process of ISTA is expanded into a deep network, and the linear inverse transformation from the received pilot signal to the CSI is solved using the expanded network. The experimental results show that the proposed algorithm improves the NMSE performance gain by about 3 dB over the GM-LAMP algorithm when the signal-to-noise ratio (SNR) is 15 dB, and the algorithm accelerates the convergence speed compared with the conventional CS channel estimation algorithm.
{"title":"Research on Beamspace Channel Estimation Method Based on DISTA","authors":"Juanyi Zheng, Yuanyuan Lv, Jinyu Mu, Lirong Xing, Pei Jie","doi":"10.1109/ICNLP58431.2023.00089","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00089","url":null,"abstract":"Since the conventional Compressed Sensing (CS) algorithm in millimeter wave Massive Multi-Input Multi-Output (MIMO) system has the problem of low channel estimation accuracy, a deep learning based beamspace channel estimation method, Deep Iterative Shrinkage-Thresholding Algorithm (DISTA), is proposed. First, due to the sparsity of the beamspace channel, the beamspace channel estimation problem can be transformed into a sparse signal recovery problem; second, based on the iterative shrinkage threshold algorithm (ISTA), the channel state information (CSI) is sparsified using nonlinear transformation functions to replace the traditional manual transformation; finally, the iterative process of ISTA is expanded into a deep network, and the linear inverse transformation from the received pilot signal to the CSI is solved using the expanded network. The experimental results show that the proposed algorithm improves the NMSE performance gain by about 3 dB over the GM-LAMP algorithm when the signal-to-noise ratio (SNR) is 15 dB, and the algorithm accelerates the convergence speed compared with the conventional CS channel estimation algorithm.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82029109","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the development of intelligent terminals and telecommunications technology, many new applications such as driverless driving,Internet of things continues to emerge, in order to meet the user's low-latency response needs, mobile edge computing (MEC) came into being. At present, mobile edge computing mainly studies how to reduce the latency and energy consumption of users, when processing tasks, in the face of some dense tasks, the ECS processing delay is too long, but the local edge server has a lot of idleness. In order to reduce latency and energy consumption, this paper proposes an edge cloud collaborative offload strategy based on improved ant colony algorithm (IACO). The final simulation results are compared with the random unloading algorithm, the local unloading algorithm and the traditional ant colony algorithm algorithm, and the improved ant colony algorithm is the effect is the best.
{"title":"Research on Collaborative Computational Offload Strategy Based on Improved Ant Colony Algorithm in Edge Computing","authors":"Haibo Ge, Jiajun Geng, Yu An, Haodong Feng, Ting Zhou, Chaofeng Huang","doi":"10.1109/icnlp58431.2023.00093","DOIUrl":"https://doi.org/10.1109/icnlp58431.2023.00093","url":null,"abstract":"With the development of intelligent terminals and telecommunications technology, many new applications such as driverless driving,Internet of things continues to emerge, in order to meet the user's low-latency response needs, mobile edge computing (MEC) came into being. At present, mobile edge computing mainly studies how to reduce the latency and energy consumption of users, when processing tasks, in the face of some dense tasks, the ECS processing delay is too long, but the local edge server has a lot of idleness. In order to reduce latency and energy consumption, this paper proposes an edge cloud collaborative offload strategy based on improved ant colony algorithm (IACO). The final simulation results are compared with the random unloading algorithm, the local unloading algorithm and the traditional ant colony algorithm algorithm, and the improved ant colony algorithm is the effect is the best.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86393415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/icnlp58431.2023.00076
Jin Wu, Haoran Feng, Xiangyang Shi, He Wen
Due to the limitation of communication environment, the change of communication distance and the attenuation and superposition of signals in the transmission process, the range of wireless signal strength received by the receiver fluctuates greatly, which seriously affects the accurate demodulation of signals by the receiver. The radio frequency receiver with Automatic Gain Control (AGC) circuit only needs a low-bit Analog-to-Digital Converter (ADC) to quantize the received signal, which reduces the accuracy requirement of ADC circuit. Moreover, image suppression and signal demodulation in digital domain can be more accurate and flexible. In this paper, by analyzing and comparing the common circuit structures of closed-loop feedback type, open-loop feedforward type and sampling data feedback type, an open-loop feedforward digital-analog hybrid AGC circuit is designed according to the application environment requirements of the wireless communication system in the Internet of Things mode, which has the advantages of fast establishment and high linearity.
{"title":"Automatic Gain Control Circuit Design for Wireless RF Receiver","authors":"Jin Wu, Haoran Feng, Xiangyang Shi, He Wen","doi":"10.1109/icnlp58431.2023.00076","DOIUrl":"https://doi.org/10.1109/icnlp58431.2023.00076","url":null,"abstract":"Due to the limitation of communication environment, the change of communication distance and the attenuation and superposition of signals in the transmission process, the range of wireless signal strength received by the receiver fluctuates greatly, which seriously affects the accurate demodulation of signals by the receiver. The radio frequency receiver with Automatic Gain Control (AGC) circuit only needs a low-bit Analog-to-Digital Converter (ADC) to quantize the received signal, which reduces the accuracy requirement of ADC circuit. Moreover, image suppression and signal demodulation in digital domain can be more accurate and flexible. In this paper, by analyzing and comparing the common circuit structures of closed-loop feedback type, open-loop feedforward type and sampling data feedback type, an open-loop feedforward digital-analog hybrid AGC circuit is designed according to the application environment requirements of the wireless communication system in the Internet of Things mode, which has the advantages of fast establishment and high linearity.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82196463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/ICNLP58431.2023.00079
Jin Wu, Wenting Pang, Haoran Feng, Zhaoqi Zhang
With the research and development of Graph Neural Networks (GNNs), GNN has shown very good results in link prediction, node classification, social network and other applications. In this paper, the node classification algorithm based on GNN is implemented by software, and the neural network models that need hardware acceleration are selected and trained. The comparative experiments are conducted on Cora, CiteSeer and PubMed citation network datasets respectively. Through the model training of the combination of different aggregation update functions, the comprehensive analysis of the experimental results shows that the combination of message passing layer functions used in this paper has the best effect, and the test accuracy in three data sets reaches 77%, 59% and 75% respectively. In order to better deploy the network model on the hardware, the symmetric quantization operation is carried out to reduce the parameters, so as to achieve the acceleration of the software part. The experimental results show that the accuracy of the quantized model is almost unchanged.
{"title":"Implementation of Node Classification Algorithm Based on Graph Neural Network","authors":"Jin Wu, Wenting Pang, Haoran Feng, Zhaoqi Zhang","doi":"10.1109/ICNLP58431.2023.00079","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00079","url":null,"abstract":"With the research and development of Graph Neural Networks (GNNs), GNN has shown very good results in link prediction, node classification, social network and other applications. In this paper, the node classification algorithm based on GNN is implemented by software, and the neural network models that need hardware acceleration are selected and trained. The comparative experiments are conducted on Cora, CiteSeer and PubMed citation network datasets respectively. Through the model training of the combination of different aggregation update functions, the comprehensive analysis of the experimental results shows that the combination of message passing layer functions used in this paper has the best effect, and the test accuracy in three data sets reaches 77%, 59% and 75% respectively. In order to better deploy the network model on the hardware, the symmetric quantization operation is carried out to reduce the parameters, so as to achieve the acceleration of the software part. The experimental results show that the accuracy of the quantized model is almost unchanged.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83351709","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/icnlp58431.2023.00085
Youyao Liu, Xinxin Liu, Kai Zhou, Qifei Shi
With the remarkable effectiveness of recurrent neural network (RNN) in speech recognition, machine translation and other fields, more and more scholars at home and abroad have begun to pay attention to the research of cyclic neural network acceleration. In recent years, due to the increase of the scale of the recurrent neural network, the software can speed up the network through the weight pruning network model compression technology. The acceleration of the cyclic neural network does not only stay in the aspect of software acceleration, but also in the aspect of hardware, the acceleration strategy includes the design of RNN accelerator based on GPU, FPGA and special ASIC circuit. The storage system almost determines the upper limit of the working efficiency of the accelerator. When the input data cannot be provided to the computing unit in time, the computing unit has to enter the idle state frequently, resulting in low working efficiency. Therefore, storage systems with continuous data feeds are very important for accelerators. This paper proposes a mapping mechanism of MVM operations on hardware operation units, and proposes a storage system with continuous data feeds.
{"title":"Design of Memory System for Recursive Neural Network Hardware Accelerator","authors":"Youyao Liu, Xinxin Liu, Kai Zhou, Qifei Shi","doi":"10.1109/icnlp58431.2023.00085","DOIUrl":"https://doi.org/10.1109/icnlp58431.2023.00085","url":null,"abstract":"With the remarkable effectiveness of recurrent neural network (RNN) in speech recognition, machine translation and other fields, more and more scholars at home and abroad have begun to pay attention to the research of cyclic neural network acceleration. In recent years, due to the increase of the scale of the recurrent neural network, the software can speed up the network through the weight pruning network model compression technology. The acceleration of the cyclic neural network does not only stay in the aspect of software acceleration, but also in the aspect of hardware, the acceleration strategy includes the design of RNN accelerator based on GPU, FPGA and special ASIC circuit. The storage system almost determines the upper limit of the working efficiency of the accelerator. When the input data cannot be provided to the computing unit in time, the computing unit has to enter the idle state frequently, resulting in low working efficiency. Therefore, storage systems with continuous data feeds are very important for accelerators. This paper proposes a mapping mechanism of MVM operations on hardware operation units, and proposes a storage system with continuous data feeds.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80590737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
LoRa (Long Range), a wireless communication technology designed for Low Power Wide Area Networks (LPWAN), facilitates diverse IoT applications and inter-device communication by virtue of its openness and adaptable network deployment. However, the conventional static link transmission scheme employed in practical LoRa network deployment fails to fully exploit the available channel resources in dynamic channel environments, resulting in suboptimal network performance. To address this issue, this paper proposes a more efficient Adaptive Data Rate (ADR) algorithm tailored for low-cost gateways. This algorithm incorporates fuzzy support vector machine (FSVM) to accurately classify link quality and employs distinct link adaptation algorithms based on varying link qualities. Notably, the algorithm considers both link-level performance and MAC layer performance. Experimental measurements demonstrate that our proposed algorithm surpasses the standard LoRaWAN ADR algorithm in terms of packet reception rate (PRR) and network throughput in both single end device (ED) and multi EDs scenarios. Specifically, in multi-EDs scenarios, the proposed algorithm yields a remarkable 34.12% improvement in throughput and a significant 26% enhancement in packet reception rate compared to the LoRaWAN ADR algorithm. These findings demonstrate the substantial enhancements achieved by the proposed algorithm in terms of network throughput and packet reception rate.
{"title":"A Collision-Reducible Adaptive Data Rate Algorithm for Low-cost LoRa Gateways","authors":"Honggang Wang, Peidong Pei, Ruoyu Pan, Lihua Jie, Ruixue Yu, Kai Wu","doi":"10.1109/ICNLP58431.2023.00087","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00087","url":null,"abstract":"LoRa (Long Range), a wireless communication technology designed for Low Power Wide Area Networks (LPWAN), facilitates diverse IoT applications and inter-device communication by virtue of its openness and adaptable network deployment. However, the conventional static link transmission scheme employed in practical LoRa network deployment fails to fully exploit the available channel resources in dynamic channel environments, resulting in suboptimal network performance. To address this issue, this paper proposes a more efficient Adaptive Data Rate (ADR) algorithm tailored for low-cost gateways. This algorithm incorporates fuzzy support vector machine (FSVM) to accurately classify link quality and employs distinct link adaptation algorithms based on varying link qualities. Notably, the algorithm considers both link-level performance and MAC layer performance. Experimental measurements demonstrate that our proposed algorithm surpasses the standard LoRaWAN ADR algorithm in terms of packet reception rate (PRR) and network throughput in both single end device (ED) and multi EDs scenarios. Specifically, in multi-EDs scenarios, the proposed algorithm yields a remarkable 34.12% improvement in throughput and a significant 26% enhancement in packet reception rate compared to the LoRaWAN ADR algorithm. These findings demonstrate the substantial enhancements achieved by the proposed algorithm in terms of network throughput and packet reception rate.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74936604","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/ICNLP58431.2023.00024
Guangyuan Zhao, Xu Huang
Traditional plant disease recognition algorithms have a complicated approach, difficult feature extraction, and low recognition accuracy. Based on the improved EfficientNetV2 model, this research classifies images of apple leaf disease. This study collected images of seven common apple leaf disease categories and one healthy category to address the present needs of various complex disease recognition scenarios. The disease images not only contain the common laboratory background but also add the background of the field growth environment of apple trees. And different recognition scenarios are further enriched by image enhancement techniques. For the model part, the processing of spatial feature information was strengthened while focusing on the channel feature information to ensure that the model focuses more on the subtle disease spot information for different disease classifications. The experimental results show that the accuracy of the model training recognition is 97.49%. To better evaluate this study, comparison experiments were conducted with five other popular convolutional neural network classification models, such as ResNet-50, DenseNet-121, Xception, MobileNet, and EfficientNet-B3. The improved models enhance the recognition accuracy of complex scenes and improve the model parameters and training speed. It provides a reference for apple leaf disease recognition and the development needs of smart agriculture.
{"title":"Apple Leaf Disease Recognition Based on Improved Convolutional Neural Network with an Attention Mechanism","authors":"Guangyuan Zhao, Xu Huang","doi":"10.1109/ICNLP58431.2023.00024","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00024","url":null,"abstract":"Traditional plant disease recognition algorithms have a complicated approach, difficult feature extraction, and low recognition accuracy. Based on the improved EfficientNetV2 model, this research classifies images of apple leaf disease. This study collected images of seven common apple leaf disease categories and one healthy category to address the present needs of various complex disease recognition scenarios. The disease images not only contain the common laboratory background but also add the background of the field growth environment of apple trees. And different recognition scenarios are further enriched by image enhancement techniques. For the model part, the processing of spatial feature information was strengthened while focusing on the channel feature information to ensure that the model focuses more on the subtle disease spot information for different disease classifications. The experimental results show that the accuracy of the model training recognition is 97.49%. To better evaluate this study, comparison experiments were conducted with five other popular convolutional neural network classification models, such as ResNet-50, DenseNet-121, Xception, MobileNet, and EfficientNet-B3. The improved models enhance the recognition accuracy of complex scenes and improve the model parameters and training speed. It provides a reference for apple leaf disease recognition and the development needs of smart agriculture.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85121251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/icnlp58431.2023.00025
Sun Wei, Sun Yi Jun, Li Zhao Chen, Guo Jing
In order to meet the urgent needs of automation and intelligent picking of kiwifruit, aiming at the problems of unreasonable construction of kiwifruit data set, low fruit recognition accuracy and poor spatial positioning in the natural environment of orchard, a precise recognition and visual positioning method of kiwifruit based on improved Yolov5s was proposed. In view of the growth characteristics of kiwifruit in trellis orchards, a multi-type kiwifruit data set was first constructed. Furthermore, the attention mechanism and multi-scale module are combined to improve the Yolov5s network structure, identify kiwifruit and extract the center coordinates of the prediction box. The experimental results show that the average accuracy of the model for six kiwifruit types under different weather and light conditions is 98 %. The single image recognition time of $1280times 720$ pixel is about 13.8 ms, and the weight is only 15.21 Mb. It can be seen that this study can provide technical support for the vision system of kiwifruit automatic picking robot, and provide reference for the intelligent recognition and positioning of other fruits (such as apples, mangoes and oranges).
{"title":"Accurate Recognition of Kiwifruit Based on Improved YOLOv5","authors":"Sun Wei, Sun Yi Jun, Li Zhao Chen, Guo Jing","doi":"10.1109/icnlp58431.2023.00025","DOIUrl":"https://doi.org/10.1109/icnlp58431.2023.00025","url":null,"abstract":"In order to meet the urgent needs of automation and intelligent picking of kiwifruit, aiming at the problems of unreasonable construction of kiwifruit data set, low fruit recognition accuracy and poor spatial positioning in the natural environment of orchard, a precise recognition and visual positioning method of kiwifruit based on improved Yolov5s was proposed. In view of the growth characteristics of kiwifruit in trellis orchards, a multi-type kiwifruit data set was first constructed. Furthermore, the attention mechanism and multi-scale module are combined to improve the Yolov5s network structure, identify kiwifruit and extract the center coordinates of the prediction box. The experimental results show that the average accuracy of the model for six kiwifruit types under different weather and light conditions is 98 %. The single image recognition time of $1280times 720$ pixel is about 13.8 ms, and the weight is only 15.21 Mb. It can be seen that this study can provide technical support for the vision system of kiwifruit automatic picking robot, and provide reference for the intelligent recognition and positioning of other fruits (such as apples, mangoes and oranges).","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90540511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-03-01DOI: 10.1109/ICNLP58431.2023.00031
Ziyu Mao, Bo Li, Lei Dong, Yani Qiao, Hao Sun, Yuji Li
In the field of array signal processing, multiple signal classification (MUSIC) algorithm is a classical spectrum estimation algorithm. However, when there are coherent signals, the rank of signal covariance matrix is generally less than the number of signals, which makes the estimation inaccurate. Taking uniform linear array (ULA) as an example, this paper presents a high-precision DOA estimation algorithm by reconstructing noise subspace. This algorithm uses not only the auto-covariance but also cross-covariance information and constructs a new augmented matrix with the auto-covariance matrix. Noise subspace and eigenvalue matrix can be obtained by singular value decomposition of matrix. For more reliable data, on the basis of a large number of experiments, a noise subspace consisting of the eigenvectors corresponding to the new eigenvalue matrix is reconstructed, and finally the DOA estimation is obtained through spectrum peak search. It is shown by the simulation results show that the improved algorithm can maintain the accuracy well of DOA with effect even under the conditions of low signal-to-noise ratio and small number of snapshots.
{"title":"An Effective Algorithm for Direction-of-Arrival Estimation of Coherent Signals with ULA","authors":"Ziyu Mao, Bo Li, Lei Dong, Yani Qiao, Hao Sun, Yuji Li","doi":"10.1109/ICNLP58431.2023.00031","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00031","url":null,"abstract":"In the field of array signal processing, multiple signal classification (MUSIC) algorithm is a classical spectrum estimation algorithm. However, when there are coherent signals, the rank of signal covariance matrix is generally less than the number of signals, which makes the estimation inaccurate. Taking uniform linear array (ULA) as an example, this paper presents a high-precision DOA estimation algorithm by reconstructing noise subspace. This algorithm uses not only the auto-covariance but also cross-covariance information and constructs a new augmented matrix with the auto-covariance matrix. Noise subspace and eigenvalue matrix can be obtained by singular value decomposition of matrix. For more reliable data, on the basis of a large number of experiments, a noise subspace consisting of the eigenvectors corresponding to the new eigenvalue matrix is reconstructed, and finally the DOA estimation is obtained through spectrum peak search. It is shown by the simulation results show that the improved algorithm can maintain the accuracy well of DOA with effect even under the conditions of low signal-to-noise ratio and small number of snapshots.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88868554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
At present, algorithms for human pose estimation and image caption are prosperous but have disadvantages. The current mainstream algorithms of pose estimation only present the information of key nodes as a scalar but lacks semantics, while in most of algorithms for human image captioning, more attention is paid to the relationship between human bodies and the background, without understanding the human body semantics, which can not meet the need of deep visual understanding.In this paper, to fill in imperfection in previous studies, we provide a novel data set of the caption of human pose estimation for the deep understanding of image semantics. Moreover, we use the pose estimation system to extract posture figures and then we utilize the encoder-decoder to generate the captions of human poses in single picture, to produce deeper understanding of the original image. Lastly, we use Bert to carry out the next step of reasoning and get a further understanding. Our data set is open source.
{"title":"What are You Posing: A gesture description dataset based on coarse-grained semantics","authors":"Luchun Chen, Guorun Wang, Yaoru Sun, Rui Pang, Chengzhi Zhang","doi":"10.1109/ICNLP58431.2023.00044","DOIUrl":"https://doi.org/10.1109/ICNLP58431.2023.00044","url":null,"abstract":"At present, algorithms for human pose estimation and image caption are prosperous but have disadvantages. The current mainstream algorithms of pose estimation only present the information of key nodes as a scalar but lacks semantics, while in most of algorithms for human image captioning, more attention is paid to the relationship between human bodies and the background, without understanding the human body semantics, which can not meet the need of deep visual understanding.In this paper, to fill in imperfection in previous studies, we provide a novel data set of the caption of human pose estimation for the deep understanding of image semantics. Moreover, we use the pose estimation system to extract posture figures and then we utilize the encoder-decoder to generate the captions of human poses in single picture, to produce deeper understanding of the original image. Lastly, we use Bert to carry out the next step of reasoning and get a further understanding. Our data set is open source.","PeriodicalId":53637,"journal":{"name":"Icon","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81862669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：