Zhiqiang Bao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhiqiang Bao has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Jinhang Su
Xi'an University of Posts & Telecommunications, China
2 shared publications - SiWei Wang
School of Communication and Information Engineering, Xi'an University of Posts & Telecommunications, Xi'an, shaanxi, China
2 shared publications - Ting Ai
Xi'an University of Posts & Telecommunications, China
2 shared publications - Bohan Zhang
School of Communication and Information Engineering, Xi'an University of Posts & Telecommunications, Xi'an, shaanxi, China
1 shared publication Chuqi Shen
Xian University of Posts and Telecommunications,School of Communication and Information Engineering,Xian,China
1 shared publication- Defei Du
School of Communication and Information Engineering, Xi'an University of Posts & Telecommunications, Xi'an, shaanxi, China
1 shared publication
Latest Publications
- Open Access
- 20 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 28-34
Small object detection algorithm based on improved YOLOv7 multi-layer attention fusion
- 20 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 28-34
Open AccessSmall object detection algorithm based on improved YOLOv7 multi-layer attention fusion
- Open Access
- 20 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 770-776
Human Parsing Network Algorithm Based on Improved YOLOv7
- 20 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 770-776
Open AccessHuman Parsing Network Algorithm Based on Improved YOLOv7
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 287-292
Facial emotion recognition based on hierarchical attention mechanism
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 287-292
Open AccessFacial emotion recognition based on hierarchical attention mechanism
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 498-505
Design of Real-Time Face Detection FPGA System Based on MTCNN
Face detection and alignment are challenging operations due to variations in image angles, background lighting conditions, and intermediate blocking objects. Recent work has shown that these tasks can be improved by using a multi-task cascaded convolutional neural network (MTCNN) architecture. However, it is difficult to implement such an approach in a low-end edge AI system due to its high computational complexity. In this paper, a low cost, low power real-time face detection system based on FPGA MTCNN is designed. Firstly, the computational requirements of the algorithm are analyzed. Based on this analysis, a compromise solution is adopted. Data accuracy is a 16-bit floating point number. Meanwhile, the dual parallel method in and between convolutional kernels is used to reduce latency, and only the output of each level network is stored in block-on-chip RAM to reduce hardware resource consumption. Experimental results show that the system can significantly improve throughput, reduce time delay and power consumption, and Compared to 8-bit quantization, the detection accuracy is improved. The system takes into account the accuracy of the algorithm and hardware resources to realize real-time processing.
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 498-505
Open AccessDesign of Real-Time Face Detection FPGA System Based on MTCNN
Face detection and alignment are challenging operations due to variations in image angles, background lighting conditions, and intermediate blocking objects. Recent work has shown that these tasks can be improved by using a multi-task cascaded convolutional neural network (MTCNN) architecture. However, it is difficult to implement such an approach in a low-end edge AI system due to its high computational complexity. In this paper, a low cost, low power real-time face detection system based on FPGA MTCNN is designed. Firstly, the computational requirements of the algorithm are analyzed. Based on this analysis, a compromise solution is adopted. Data accuracy is a 16-bit floating point number. Meanwhile, the dual parallel method in and between convolutional kernels is used to reduce latency, and only the output of each level network is stored in block-on-chip RAM to reduce hardware resource consumption. Experimental results show that the system can significantly improve throughput, reduce time delay and power consumption, and Compared to 8-bit quantization, the detection accuracy is improved. The system takes into account the accuracy of the algorithm and hardware resources to realize real-time processing.
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 337-344
Research on CNN-based running arm swing action recognition method
Correct arm swing posture helps improve overall running performance during the running process. This article analyzes incorrect arm swing postures such as Over-swing arms, Crossover Arm swing, and Abduction of the arm in running. It designs and develops a running arm swing monitoring system. The system utilizes wearable devices to collect data, preprocess data, and calculate arm swing posture angles using AEKF. The data is then sent to the running arm swing monitoring platform's host computer through a WIFI wireless communication module for windowing processing, time-domain feature extraction, and arm swing posture recognition. Additionally, this article performs algorithm performance analysis on CNN and selects the best iterative algorithm and wearing position. By comparing with the performance of traditional machine learning algorithms, the effectiveness of this method is demonstrated. The classification accuracy of the system is above 93%, further validating its reliability and accuracy.
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 337-344
Open AccessResearch on CNN-based running arm swing action recognition method
Correct arm swing posture helps improve overall running performance during the running process. This article analyzes incorrect arm swing postures such as Over-swing arms, Crossover Arm swing, and Abduction of the arm in running. It designs and develops a running arm swing monitoring system. The system utilizes wearable devices to collect data, preprocess data, and calculate arm swing posture angles using AEKF. The data is then sent to the running arm swing monitoring platform's host computer through a WIFI wireless communication module for windowing processing, time-domain feature extraction, and arm swing posture recognition. Additionally, this article performs algorithm performance analysis on CNN and selects the best iterative algorithm and wearing position. By comparing with the performance of traditional machine learning algorithms, the effectiveness of this method is demonstrated. The classification accuracy of the system is above 93%, further validating its reliability and accuracy.
- Open Access
- 24 March 2023
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Millimeter Wave Radar Fall Detection Algorithm Based on Improved Transformer
Aiming at the defects of convolutional neural network that it is difficult to extract high-level visual semantic information and ignore inter-channel information, a millimeter wave radar fall detection algorithm based on improved Transformer is proposed. By combining the channel attention mechanism with the Transformer network structure to form a pyramid structure, the temporal information and spatial information of the signal are effectively extracted, the feature extraction ability of the deep learning network model is enhanced, and the problem of overfitting of the Transformer structure under small samples is improved. The fall detection of millimeter wave radar signal is realized. The experimental results show that the classification accuracy of the algorithm is 96.8%, which verifies the feasibility and effectiveness of the model.\
- 24 March 2023
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Open AccessMillimeter Wave Radar Fall Detection Algorithm Based on Improved Transformer
Aiming at the defects of convolutional neural network that it is difficult to extract high-level visual semantic information and ignore inter-channel information, a millimeter wave radar fall detection algorithm based on improved Transformer is proposed. By combining the channel attention mechanism with the Transformer network structure to form a pyramid structure, the temporal information and spatial information of the signal are effectively extracted, the feature extraction ability of the deep learning network model is enhanced, and the problem of overfitting of the Transformer structure under small samples is improved. The fall detection of millimeter wave radar signal is realized. The experimental results show that the classification accuracy of the algorithm is 96.8%, which verifies the feasibility and effectiveness of the model.\
- Open Access
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 657-663
Radar micro moving gesture recognition method based on multi-scale fusion deep network
In order to solve the problem that the micro moving gesture features are not obvious and difficult to be identified, a micro moving gesture recognition method based on multi-scale fusion deep network for millimeter wave radar is proposed in this paper. The method is mainly composed of 2D convolution module, multi-scale fusion module and attention mechanism module. The multi-scale fusion module is composed of three residual blocks of different scales, which can obtain receptive fields of different sizes and obtain multi-scale features. Meanwhile, residual blocks of different scales are fused to increase the diversity of the network and better extract the deep features of the data. The Squeeze-and-congestion (SE) attention mechanism module is added to suppress the channel characteristics with little information. This improves the network identification accuracy and reduces the number of parameters and computation. The experimental results show that this method is simple to implement, doesn't need to do complex data preprocessing. The convergence speed of the network is fast, which can realize the effective recognition of the micro moving gesture.
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 657-663
Open AccessRadar micro moving gesture recognition method based on multi-scale fusion deep network
In order to solve the problem that the micro moving gesture features are not obvious and difficult to be identified, a micro moving gesture recognition method based on multi-scale fusion deep network for millimeter wave radar is proposed in this paper. The method is mainly composed of 2D convolution module, multi-scale fusion module and attention mechanism module. The multi-scale fusion module is composed of three residual blocks of different scales, which can obtain receptive fields of different sizes and obtain multi-scale features. Meanwhile, residual blocks of different scales are fused to increase the diversity of the network and better extract the deep features of the data. The Squeeze-and-congestion (SE) attention mechanism module is added to suppress the channel characteristics with little information. This improves the network identification accuracy and reduces the number of parameters and computation. The experimental results show that this method is simple to implement, doesn't need to do complex data preprocessing. The convergence speed of the network is fast, which can realize the effective recognition of the micro moving gesture.
- Open Access
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 180-186
A Joint Model for Text Classification with BERT-BiLSTM and GCN
With the development of Graph Neural Network (GNN), a lot of GNN-based methods have been proposed in text classification. However, GNNs are difficult to capture the word relationship of the context in the document sequence. In addition, these models are common to use one-hot encoding to initialize the node features, which have high dimensions and sparse features. Therefore, it is difficult to represent the correlation between texts. To effectively learn the semantic features of contexts and the association between words, we propose a text classification method based on the BERT-BiLSTM word embedding model and the Graph Convolutional Network (GCN), which is named BBG. The BERT-BiLSTM can learn word embedding representations in different contexts and the semantic relationship of text contexts. The learned representation vectors are provided into GCN as the features of document nodes. The feature representations of the training data and unlabeled test data are learned by performing the label propagation in GCN. Finally, the classification results are output by utilizing the Softmax layer. Experiments on four benchmark datasets show that our method has better performances than existing text classification methods.
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 180-186
Open AccessA Joint Model for Text Classification with BERT-BiLSTM and GCN
With the development of Graph Neural Network (GNN), a lot of GNN-based methods have been proposed in text classification. However, GNNs are difficult to capture the word relationship of the context in the document sequence. In addition, these models are common to use one-hot encoding to initialize the node features, which have high dimensions and sparse features. Therefore, it is difficult to represent the correlation between texts. To effectively learn the semantic features of contexts and the association between words, we propose a text classification method based on the BERT-BiLSTM word embedding model and the Graph Convolutional Network (GCN), which is named BBG. The BERT-BiLSTM can learn word embedding representations in different contexts and the semantic relationship of text contexts. The learned representation vectors are provided into GCN as the features of document nodes. The feature representations of the training data and unlabeled test data are learned by performing the label propagation in GCN. Finally, the classification results are output by utilizing the Softmax layer. Experiments on four benchmark datasets show that our method has better performances than existing text classification methods.
- Open Access
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 939-944
Improved atrial fibrillation recognition algorithm based on residual network
An improved residual network model is proposed to deal with the complex and changeable characteristics of one-dimensional electrocardiogram. In this model, firstly, in order to avoid the network degradation problem of the model along with the deepening of the number of layers, when extracting various deep-level features of ECG signals using multiple convolution layers in CNN, the residual module is integrated into the network, and an appropriate shortcut connection is selected to connect the input with the superposition output of the corresponding convolution layer to construct a deep residual network to extract more abstract signal features. Secondly, the output of the last residual module is sent to the GAP layer, and the parameters of this layer are greatly reduced compared with those of the full connection layer, which is equivalent to the compression of the model, and thus the over-fitting of the model is avoided to a certain extent. Finally, the original ECG signals were automatically classified based on the PCinCC2017 database to complete the recognition of atrial fibrillation. Experimental results show that the proposed algorithm has a classification accuracy of 86% and a F1 measure of 83%, which prove the feasibility of the model and the effectiveness of the algorithm.
- 23 September 2022
- Published by Association for Computing Machinery (ACM)
- p. 939-944
Open AccessImproved atrial fibrillation recognition algorithm based on residual network
An improved residual network model is proposed to deal with the complex and changeable characteristics of one-dimensional electrocardiogram. In this model, firstly, in order to avoid the network degradation problem of the model along with the deepening of the number of layers, when extracting various deep-level features of ECG signals using multiple convolution layers in CNN, the residual module is integrated into the network, and an appropriate shortcut connection is selected to connect the input with the superposition output of the corresponding convolution layer to construct a deep residual network to extract more abstract signal features. Secondly, the output of the last residual module is sent to the GAP layer, and the parameters of this layer are greatly reduced compared with those of the full connection layer, which is equivalent to the compression of the model, and thus the over-fitting of the model is avoided to a certain extent. Finally, the original ECG signals were automatically classified based on the PCinCC2017 database to complete the recognition of atrial fibrillation. Experimental results show that the proposed algorithm has a classification accuracy of 86% and a F1 measure of 83%, which prove the feasibility of the model and the effectiveness of the algorithm.
- 1 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 310-314
Research on Sleep Staging Method Based on DeepStageNet
In order to solve the problem of inaccurate results in the prediction of sleep stages by traditional networks, a DeepStageNet network was proposed. The network firstly express that deep characteristics through three-layer parallel CNN. Secondly, the extracted features are respectively transferred to the density at the next layer and further backward to the BiLSTM, which continues to transfer information and mine the dependency relationship between the various stages of sleep in the same way, thus realizing the automatic learning of the staging features of EEG signals and the determination of sleep periods. Finally, the experimental results on Sleep-EDF Database show that, after removing the influence of the data in the half-hour awake period before and after sleep, the Staging Accuracy of DeepStageNet Network on the traditiona130s data set and the new 210s data set after secondary processing are 86.5% and 87.7%. This effectively improves the Staging Accuracy of Sleep Model, and has certain practical value.
- 1 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 310-314
Research on Sleep Staging Method Based on DeepStageNet
In order to solve the problem of inaccurate results in the prediction of sleep stages by traditional networks, a DeepStageNet network was proposed. The network firstly express that deep characteristics through three-layer parallel CNN. Secondly, the extracted features are respectively transferred to the density at the next layer and further backward to the BiLSTM, which continues to transfer information and mine the dependency relationship between the various stages of sleep in the same way, thus realizing the automatic learning of the staging features of EEG signals and the determination of sleep periods. Finally, the experimental results on Sleep-EDF Database show that, after removing the influence of the data in the half-hour awake period before and after sleep, the Staging Accuracy of DeepStageNet Network on the traditiona130s data set and the new 210s data set after secondary processing are 86.5% and 87.7%. This effectively improves the Staging Accuracy of Sleep Model, and has certain practical value.