UROP Research Mentor Project Submission Portal: Submission #721
Submission information
Submission Number: 721
Submission ID: 14231
Submission UUID: 0751e183-262f-4184-a9b5-172f0a95e5cd
Submission URI: /urop-research-mentor-project-submission-portal
Submission Update: /urop-research-mentor-project-submission-portal?token=_-PVSAd9AjxVzB8bkYhfi9yCyRkHd5SpUi7yYqFnSnM
Created: Wed, 08/07/2024 - 04:18 PM
Completed: Wed, 08/07/2024 - 05:26 PM
Changed: Thu, 10/10/2024 - 08:50 AM
Remote IP address: 128.186.121.249
Submitted by: Anonymous
Language: English
Is draft: No
Webform: UROP Project Proposal Portal
Submitted to: UROP Research Mentor Project Submission Portal
Research Mentor Information
Additional Research Mentor(s)
Overall Project Details
Action Recognition based on Human Mesh Sequences
Action Recognition, Deep Learning, RF Sensing
No
2
Computer Science, Electrical Engineering
On FSU Main Campus
No, the project is remote
Partially Remote
10
Flexible schedule (Combination of business and outside of business. TBD between student and research mentor.)
Other than previous approaches, instead of using Wi-Fi signals to generate a human skeleton sequence, we would generate a human mesh sequence and use it for action recognition. Using human mesh sequence as raw data will bring more information compared with using skeleton sequence, for instance, body surface, torso rotation, etc., which would improve the accuracies of the final action recognition. On the other hand, due to the significant increase in data size, mesh based deep learning model will require much more computational resources. Thus, an efficient down-sampling methodology becomes critical in this project.
Research Tasks:
a. Literature Review:
1. Past papers survey
2. Sampling methods they have used
3. Data categories they have used
4. Deep learning models
b. Data collection: Both RGB and RF data, total 868 data clips with 13 persons
c. Data analysis:
1. Robustness of the RF based mesh (Wi-Mesh)
2. Effectiveness of different down-sampling methods
3. Deep learning results of mesh sequences with different numbers of vertices
a. Literature Review:
1. Past papers survey
2. Sampling methods they have used
3. Data categories they have used
4. Deep learning models
b. Data collection: Both RGB and RF data, total 868 data clips with 13 persons
c. Data analysis:
1. Robustness of the RF based mesh (Wi-Mesh)
2. Effectiveness of different down-sampling methods
3. Deep learning results of mesh sequences with different numbers of vertices
Programming skills in C or Python are required.
An interest in developing wireless signal processing and machine learning algorithms is required.
An interest in developing wireless signal processing and machine learning algorithms is required.
I deeply cherish the role of being an educator, taking immense pride in sharing my knowledge and inspiring the next generation. With an interdisciplinary background in both EE and CS, I bring a unique perspective to the students. This diverse knowledge base enables me to bridge the gap between EE and CS concepts, fostering a holistic understanding for students from either domain. Such a foundation not only enriches their academic journey but also prepares them to be innovators and leaders in their respective fields. I fostered an environment that facilitated a deeper and more intuitive understanding of research concepts. I dedicated months to collaborative brainstorming on paper topics, system designs, experiment conduction, and paper writing. To further reinforce the students' understanding, I sometimes posed challenging tasks, the solutions to which I already knew. Far from discouraging them, these tasks sparked their curiosity and enthusiasm, often leading them to invest more time than anticipated out of sheer excitement for creating something valuable.
https://emdcyy.github.io/
{Empty}
No
{Empty}
UROP Program Elements
Yes
Yes
Yes
Yes
{Empty}
2024
https://cre.fsu.edu/urop-research-mentor-project-submission-portal?element_parents=elements/research_mentor_information/headshot_optional_&ajax_form=1&_wrapper_format=drupal_ajax&token=_-PVSAd9AjxVzB8bkYhfi9yCyRkHd5SpUi7yYqFnSnM