Zhen Gao

Prof. Dr. Zhen Gao

McMaster University, Hamilton, Ontario, Canada
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 January 2023 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
CTChongben Tao
Chongben Tao
JCJiecheng Cao
Jiecheng Cao
CWChen Wang
Chen Wang
ZZZufeng Zhang
Zufeng Zhang
ZGZhen Gao
Zhen Gao
Current monocular 3D object detection algorithms generally suffer from inaccurate depth estimation, which leads to reduction of detection accuracy. The depth error from image-to-image generation for the stereo view is insignificant compared with the gap in single-image generation. Therefore, a novel pseudo-monocular 3D object detection framework is proposed, which is called Pseudo-Mono. Particularly, stereo images are brought into monocular 3D detection. Firstly, stereo images are taken as input, then a lightweight depth predictor is used to generate the depth map of input images. Secondly, the left input images obtained from stereo camera are used as subjects, which generate enhanced visual feature and multi-scale depth feature by depth indexing and feature matching probabilities, respectively. Finally, sparse anchors set by the foreground probability maps and the multi-scale feature maps are used as reference points to find the suitable initialization approach of object query. The encoded visual feature is adopted to enhance object query for enabling deep interaction between visual feature and depth feature. Compared with popular monocular 3D object detection methods, Pseudo-Mono is able to achieve richer fine-grained information without additional data input. Extensive experimental results on the datasets of KITTI, NuScenes, and MS-COCO demonstrate the generalizability and portability of the proposed method. The effectiveness and efficiency of Pseudo-Mono have been demonstrated by extensive ablation experiments. Experiments on a real vehicle platform have shown that the proposed method maintains high performance in complex real-world environments.

18 March 2020 in IEEE Access
Verified
Journal Article
Open access
CTChongben Tao
Chongben Tao
ZGZhen Gao
Zhen Gao
JYJinli Yan
Jinli Yan
CLChunguang Li
Chunguang Li
GCGuozeng Cui
Guozeng Cui
During the construction of indoor environmental semantic maps by robot Vision SLAM (VSLAM), there exist some problems such as low label classification accuracy and low precision under the situation of sparse feature points. In this case, this paper proposes an indoor three-dimensional semantic VSLAM algorithm based on Mask Regional Convolutional Neural Network (RCNN). Firstly, an Oriented FAST and a Rotated BRIEF (ORB) algorithms are used to extract image feature points. Secondly, a Random Sample Consensus (RANSAC) algorithm is employed to eliminate mismatched points and estimate camera position-pose changes. Then, a Mask RCNN algorithm is applied to make partial adjustments to its hyper parameter. A self-made data set is used to transfer learning, fulfilling real-time target detection and instance segmentation of a scene. A three-dimensional semantic map is constructed in combination with VSLAM algorithm. The semantic information in the environment not only improves the accuracy of VSLAM construction and positioning, but also reduces the impact of object movement on the construction by marking movable objects. Meanwhile, the VSLAM algorithm is used to calculate the positional constraints between objects and improve the accuracy of semantic understanding. Finally, by comparing with other methods, it demonstrates that this method is more correct and effective. It was also verified that the proposed method can accurately interpret the semantic information in environment for the construction of three-dimensional semantic maps.

25 January 2022 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
CTChongben Tao
Chongben Tao
JXJie Xue
Jie Xue
ZZZufeng Zhang
Zufeng Zhang
ZGZhen Gao
Zhen Gao
In this brief, a parallel Deep Deterministic Policy Gradient (DDPG) algorithm is presented for biped robot gait control. Biped robot gait control is a high-dimensional continuous problem. It is challenging to obtain a fast and stable gait. Traditional methods cannot fully utilize autonomous exploration capability of a biped robot. A multiple Actor-Critic (AC) network is established to expand the scope of exploration and improve training efficiency. For optimizing experience replay mechanism, an experience filtering unit is introduced, and a cosine similarity method is used to classify experience. Then, a Markov Decision Process (MDP) model based on knowledge and experience is designed to solve the problem of sparse rewards. Finally, experimental results show that the parallel DDPG algorithm can make the biped robot walk more quickly and stably, and the speed reaches 0.62 m/s.

17 January 2023 in IEEE Transactions on Artificial Intelligence
Verified
Journal Article
Closed access
JCJieCheng Cao
JieCheng Cao
CTChongben Tao
Chongben Tao
ZZZufeng Zhang
Zufeng Zhang
ZGZhen Gao
Zhen Gao
XLXizhao Luo
Xizhao Luo
YZYuan Zhu
Yuan Zhu
Current point-voxel fusion methods for 3D object detection could not make full use of complementary information in the field of autonomous driving. Therefore, a novel two-stage 3D object detection method, called Accelerating Point-Voxel Representation (APVR), is proposed. The advantages of Point-based feature and Voxel-based feature can be integrated into a single 3D representation. Thereby, the proposed method retains more fine-grained information of an object while maintaining high efficiency. Specifically, computational cost is reduced by adding offsets to query neighboring voxels of key-points. More fine-grained information can be obtained by calculating the matching probability between neighbouring voxels and key-points. During the optimization of the prediction boxes, virtual grid points are set to capture the spatial information between key-points. The constraint of minimum enclosing rectangle is also added to optimize the directions of the prediction boxes. A large number of experiments on the KITTI, NuScenes and Waymo datasets demonstrate great generalizability and portability of the proposed approach. The effectiveness and efficiency of APVR has been proved by comparisons with the state-of-art methods. APVR makes the real-time processing frame rate reach 40.4 Hz while ensuring high prediction accuracy.

22 December 2022 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
CTChongben Tao
Chongben Tao
SFShiping Fu
Shiping Fu
CWChen Wang
Chen Wang
XLXizhao Luo
Xizhao Luo
HLHuayi Li
Huayi Li
SZSifa Zheng
Sifa Zheng
Current 3D object detection technology for autonomous driving usually cannot efficiently utilize local sensitive points. Meanwhile, contextual feature extracted from a object is not sufficient, which easily leads to deteriorated detection accuracy of the final object estimation. For the problems, a point-voxel-based 3D dynamic object detection algorithm is proposed. Firstly, local points are grouped with a camera frustum. Then the global feature extracted by the submanifold 3D voxel CNNs is aggregated into frustum key points. Secondly, a module of vector pool with feature aggregation is used to aggregate multi-scale features of the point cloud. Moreover, the frustum raw feature and BEV feature are used for feature extension. Subsequently, the fine multi-scale feature extracted from the point cloud is used as input to a subsequent fully convolutional network for final classification and continuous estimation of oriented 3D boxes. The proposed method was compared with other state-of-the-art algorithms on the KITTI, Waymo, and nuScenes datasets. Experimental results showed that the proposed algorithm was better in accuracy, robustness, and generalization capabilities in 3D dynamic object detection. Experiments on a real scenario and extensive ablation studies also demonstrated that the proposed algorithm not only effectively controls computational cost, but also achieved more efficient results in 3D object detection.

1 November 2024 in Bioengineering
Verified
Journal Article
Open access
PDF file
YWYiping Wang
Yiping Wang
MWMustafaa Wahab
Mustafaa Wahab
THTianqi Hong
Tianqi Hong
KMKyle Molinari
Kyle Molinari
GGGail Gauvreau
Gail Gauvreau
Qiyin FangQiyin Fang
Qiyin Fang
Chronic cough is associated with several respiratory diseases and is a significant burden on physical, social, and psychological health. Non-invasive, real-time, continuous, and quantitative monitoring tools are highly desired to assess cough severity, the effectiveness of treatment, and monitor disease progression in clinical practice and research. There are currently limited tools to quantitatively measure spontaneous coughs in daily living settings in clinical trials and in clinical practice. In this study, we developed a machine learning model for the detection and classification of cough sounds. Mel spectrograms are utilized as a key feature representation to capture the temporal and spectral characteristics of coughs. We applied this approach to automate cough analysis using 300 h of audio recordings from cough challenge clinical studies conducted in a clinical lab setting. A number of machine learning algorithms were studied and compared, including decision tree, support vector machine, k-nearest neighbors, logistic regression, random forest, and neural network. We identified that for this dataset, the CRNN approach is the most effective method, reaching 98% accuracy in identifying individual coughs from the audio data. These findings provide insights into the strengths and limitations of various algorithms, highlighting the potential of CRNNs in analyzing complex cough patterns. This research demonstrates the potential of neural network models in fully automated cough monitoring. The approach requires validation in detecting spontaneous coughs in patients with refractory chronic cough in a real-life setting.