Academic Projects
4. Research Assistant (2016 - Present)
Telecommunication Department, Faculty of Electric and Electronics, HCMC University of Technology
4.1. [.PDF] [Submitted] H. T. Pham, T. C. Huynh, P. V. Bui, and H. H. Kha, “Automatic feature extraction for vietnamese sign language recognition using support vector machine,” in 2018 2nd International Conference on Recent Advances in Signal Processing, Telecommunications & Computing (SigTelCom) (SigTelCom 2018), Ho Chi Minh City, Vietnam, Jan. 2018.
Abstract: This paper aims at finding an automatic approach for extracting features of the Vietnamese sign language to classify both static Vietnamese alphabet letters and their combing diacritic marks as dynamic hand gestures. A Vietnamese sign language recognition system (VSLRS) collects all images including depth images, RGB images, and skeletal join maps to extract the desired features of each hand gesture and their own movements. These characteristics are normalized, converted to build a full Vietnamese sign language combing diacritic marks. The primary features of this system are automatically extracting the hand gestures of the observed person before the Kinect device version 1, and both dynamic and static diacritic marks are able to be recognized because of the movement detection method. Multi-classes SVM and the One-Against-All approach are employed to find two suitable SVMs for static and dynamic hand gesture recognition. During a recognition phase, all hand gestures are extracted, normalized, and then they will be filtered out based on the Euclid distance difference of hand positions in captured frames to go through the exact SVMs, the recognized letter or diacritic is the positive label of all the SVM classes. The experimental results demonstrate the proposed VSLRS having the high accuracy of VSL recognition in real-time.
Background: Support Vector Machine, one-against-all, hand gesture extraction, machine learning, Matlab programming.
Figure 4: Vietnamese Sign Language (HCMC Sign Language based on American Sign Language) alphabet letters combing diacritic marks and the VSLRS.
Figure 5: Static and dynamic right-hand gesture image extraction.
Figure 6: Experimental results of static and dynamic right-hand gestures recognition.
3. Master student research position (January 2015 - January 2016)
Telecommunication Department, Faculty of Electric and Electronics, HCMC University of Technology
[.PDF] H. T. Pham and H. H. Kha, "An efficient star skeleton extraction for human action recognition using hidden markov models," in 2016 IEEE Sixth International Conference on Communications and Electronics (ICCE) (IEEE ICCE 2016), Ha Long Bay, Vietnam, July 2016.
Abstract: This paper aims at finding an efficient approach for automatic human action recognition to classify human actions in both outdoor and indoor environments. A human action recognition system (HARS) collects video frames of human activities, extracts the desired features of each human skeleton. These characteristics are calculated, classified to build a skeleton database that can distinguish almost human gestures. This HARS converts every sequence of human gestures to the sequences of skeletal joint mapping (SJM). Then it assigns corresponding observation symbols to each SJM. Those observation sequences are used to train of hidden Markov models (HMMs) corresponding to seven actions: standing, walking, running, jumping, falling, lying, and sitting. Baum-Welch and forward-backward algorithms are employed to find optimal parameters of each HMM. During a recognition phase, each human gesture sequence is converted to an observation sequence and put into seven optimized HMM models. The current action can be identified by finding a model with the highest probability. The experimental results show that the proposed HARS offers high accuracy of action recognition in real-time.
Background: machine learning, hidden Markov model, human action recognition, digital image processing, C#-python programming.
Figure 2: human action recognition schema and 60-SJM database.
Figure 3: Seven experimental actions and its sequences.
Demonstration video [.Youtube]
2. Student research position (February 2013 - January 2014)
Research Project: "Design a RF circuit to hold the low-frequency difference of two high-frequency VCOs"
Abstract: Design the RF circuit with PLL, mixer, low-pass filter, comparison, and algorithm blocks in ADS software to hold 100Hz-difference between 2 VCOs 19.2 MHz for goal: down-sample before applying these signals into Ground Penetrate Radar (GPR).
Figure 1: RF circuit design and its experimental images.
1. Collaborator of Electric-Telecom Lab (January 2013 - March 2013)
Electric-Telecom Lab, DCSELab, HCMC University of Technology
- Collaborator of Electric-Telecom Lab, learn an introduction of Ground Penetration Radar (GPR).
- Understand the basic knowledge of the definition, operation, structure of GPR.