Trong Nguyen Nguyen

Dr. Trong Nguyen Nguyen

orcid

Research Focus & Metrics

Academic Career Length

13Years (2014-Present)

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

Representative Publications

View All Publications
26 October 2016 in Sensors
Verified
Journal Article
Open access
PDF file
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
HHHuu Hung HUYNH
Huu Hung HUYNH
JMJean Meunier
Jean Meunier
Human gait analysis plays an important role in musculoskeletal disorder diagnosis. Detecting anomalies in human walking, such as shuffling gait, stiff leg or unsteady gait, can be difficult if the prior knowledge of such a gait pattern is not available. We propose an approach for detecting abnormal human gait based on a normal gait model. Instead of employing the color image, silhouette, or spatio-temporal volume, our model is created based on human joint positions (skeleton) in time series. We decompose each sequence of normal gait images into gait cycles. Each human instant posture is represented by a feature vector which describes relationships between pairs of bone joints located in the lower body. Such vectors are then converted into codewords using a clustering technique. The normal human gait model is created based on multiple sequences of codewords corresponding to different gait cycles. In the detection stage, a gait cycle with normality likelihood below a threshold, which is determined automatically in the training step, is assumed as an anomaly. The experimental results on both marker-based mocap data and Kinect skeleton show that our method is very promising in distinguishing normal and abnormal gaits with an overall accuracy of 90.12%.

9 July 2018 in IEEE Access
Verified
Journal Article
Open access
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
HHHuu-Hung Huynh
Huu-Hung Huynh
JMJean Meunier
Jean Meunier
In order to extract more detailed features, many recent practical applications work with 3D models instead of 2D images. However, 3D reconstruction usually requires either multiple cameras or a depth sensor and a turntable. This paper proposes an approach for performing a 3D reconstruction using only one depth camera together with 2 or more mirrors. Mirrors are employed as virtual depth cameras placed at different positions. All measured depth data are provided in only one frame at each time. Significant depth distortion behind a mirror, which occurred with a standard time-of-flight (ToF) depth sensor, is reduced by removing unreliable points and/or re-estimating better positions for these points. The experiments on easyto- evaluate geometric objects show that the proposed approach could play a basic role in reconstructing intermediate 3D object models in practical applications using only cheap devices.

1 December 2014
Verified
Conference Paper
Closed access
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
DVDuc-Hoang Vo
Duc-Hoang Vo
HHHuu-Hung Huynh
Huu-Hung Huynh
JMJean Meunier
Jean Meunier
Static hand gesture recognition plays an important role in developing a system for human-computer interaction. Besides, such systems can be also used by the deaf community in order to convey information through gestures instead of words. A vision-based processing of hand gesture recognition consists of three main stages: preprocessing, feature extraction and identification. In this paper, the first stage involves following two sub-stages: segmentation which locates hand using color information and extracts its silhouette; separation that separates arm, the part with less information, based on geometrical properties. In the second stage, features which extracted from hand-without-arm are general (ratio of width to height, wrist angle and number of fingers) and detailed (calculated based on fingertips and cross sections) characteristics. Finally, support vector machine model with "max-wins" voting strategy is used to classify the hand gestures. The experiment is conducted on color image dataset of Polish Ministry of Science and Higher Education, with 89.5% classification accuracy.

8 October 2021 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
SRSebastien Roy
Sebastien Roy
JMJean Meunier
Jean Meunier
Anomaly detection is a key functionality in various vision systems, such as surveillance and security. In this work, we present a convolutional neural network (CNN) that supports the detection of anomaly, which has not been defined when building the model, at frame level. Our CNN, named SmithNet, is structured to simultaneously learn commonly occurring textures and their corresponding motion. Its architecture is a combination of: 1) an encoder extracting motion-texture coherence from each video frame and 2) two decoders that separately reconstruct the input as well as predict its typical motion from the estimated coherence. We also introduce an encoding block, which is specifically designed for the task of anomaly detection. The optimization is performed on only data of normal events, and the network is expected to determine the ones that are unusual, i.e., have not been seen before. According to the experiments on eight benchmark datasets of different environments with various anomalous events, the performance of our network is competitive or outperforms current state-of-the-art approaches.

7 February 2019 in Pattern Analysis and Applications
Verified
Journal Article
Closed access
PDF file
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
JMJean Meunier
Jean Meunier

28 May 2019 in EURASIP Journal on Image and Video Processing
Verified
Journal Article
Open access
PDF file
TNTrong Nguyen Nguyen
Trong Nguyen Nguyen
JMJean Meunier
Jean Meunier
This paper proposes a method estimating an index that indicates human gait normality based on a sequence of 3D point clouds representing the walking motion of a subject. A cylinder-based histogram is extracted from each cloud to reduce the number of data dimensions as well as highlight gait-related characteristics. A model of deep neural network is finally formed from such histograms of normal gait patterns to provide gait normality indices supporting gait assessment tasks. The ability of our approach is demonstrated using a dataset of 9 different gait types performed by 9 subjects and two other datasets converted from mocap data. The experimental results are also compared with other related methods that process different input data types including silhouette, depth map, and skeleton as well as state-of-the-art deep learning approaches working on point cloud.