School of Science and Technology 科技學院
Computing Programmes 電腦學系
| Programme | Bachelor of Science with Honours in Internet Technology |
| Supervisor | Dr. Andrew Lui |
| Areas | E-Health and Medical Applications |
| Year of Completion | 2014 |
This project aims to develop the computer respirator prediction system by finding out the key human facial features that affect the fitness of respirator. The current respirator fitting-test methodology is not efficient and it wastes a lot of resources. To select the best fitting respirator, wearers have to test respirators one by one. Those respirators fail the fitting test are wasted. A computer respirator prediction system will improve the efficiency and effectiveness of the current fitting test. The main goal of this project is to identify the key features on a human face that most likely to affect the fitness of a respirator. This project has defined a number of sub-objectives as follows:
To improve the performance of the current respirator prediction model, this project will mainly focus on finding out the key facial features affects the fitness of respirator. Machine learning algorithms are adopted to assist the selection of facial features.
A facial feature can be any identification on the human face. It can be length, angle, curvature or shape. Those information requires certain landmark points to calculate. Landmark points can be located by comparing the points' coordinate information base on the location knowledge.
Most of the N95 respirator cover half face below the eyes so it can assume that the facial points above eyes do not affect the fitness of respirator. By observing the fitness of different types of N95 respirators, the fulcrum and the contact points between the respirator and human face are likely to be related to its fitness. Nose, chin and both side of cheeks are the four main contact points between the respirator and the human face.
The contact point on the nose is a point between the nose tip and nose root. The position of that point is highly related to the shape of nose and the distance between chin and nose tip. To define the shape of nose, it can be done by measuring the vertical angle of nose, the horizontal angle of nose, the nose tall, the nose width, the nose length, the nose height, and the nose width ratio.
The nose vertical angle is the cross angle between the vector from nose tip to nose root and the vector from nose tip to subnasale. The nose horizontal angle is the cross angle between the vector from nsoe tip to left nose ala and the vector from nose tip to right nose ala. The nose tall is the z value difference between subnasale and nose tip. The nose width is the x value difference between left and right nose alas. The nose length is the distance between the nose tip and the nose root. The nose height is the y difference between the nose tip and the nose root. The nose width ratio is the ratio of the nose width to the face width.
The fulcrum and the contact point on cheeks is likely to be an important point related to the respirator fitness. A research from Texas Tech University found that the leakage of a respirator is likely to be on the area between the cheek and chain and the area between the cheek and nose tip. Those area are related to the shape of bottom half of the face. It is can be defined by the half face height, the face height, the face width, and the chin to cheek distance along the face.
The half face height is the y difference between the tip of nose and the chin. The face height is the y difference between the nose root and the chin. The face width is the x difference between both sides of cheeks. The chin to cheek distance along the face can be calculated by first adopting A* Algorithm to find the closest path from chin to cheek then calculate the length of the path and calculate the average length of both side of chin to cheeks paths.
After extract sets of facial features from the 3D model of all participants. The next step is to calculate the scale of the extracted facial features so that the actual value of all facial features can be extracted. The scale is calculated by first compare the nose width and the nose height of each participant from two set of data: the manual measurements' data and the extracted facial features data. Then, average the scale of all participants. The algorithm is:
After that, the nose tall, the nose width, the nose length, the nose height, the half face height, the face height, the face width, and the chin to cheek distance along the face are multiple with the scale as those features are calculated based on length. Key facial features can be found by adopting machine learning technique. The first thing to do is to select a classifier that produce the best prediction performance to the respirator fitness from a set of all extracted facial features. After that, the selected classifier is used to find out the key facial features by comparing each combination of facial features and the fitting test result.
| Experiment to find out key facial features for 3M 1860s | ||
|---|---|---|
| Facial Features' combination set | Precision | Conclusion |
| Full set of facial features | 64% | — |
| Chin to cheek average distance | 75% (+11%) | Respirator is likely to be fitted when > 9.13 cm |
| Experiment to find out key facial features for 3M 1862 | ||
|---|---|---|
| Facial Features' combination set | Precision | Conclusion |
| Full set of facial features | 64% | — |
| Half face height | 64% (+0%) | Respirator are likely to be fitted when > 6.5 cm |
| Experiment to find out key facial features for KCS | ||
|---|---|---|
| Facial Features' combination set | Precision | Conclusion |
| Full set of facial features | 54% | — |
| Vertical Nose Angle | 69% (+15%) | Respirator are likely to be fitted when <= 110° |
| Nose width to face width ratio | 59% (+5%) | Respirator are likely to be fitted when <= 37% |
The facial features auto-extraction algorithm extract 11 facial features which is measured base on 7 landmarks on the human face. The facial features that defines the shape of the nose are the nose height, the nose width, the nose length, the nose tall, the nose vertical angle, the nose horizontal angle and the nose width to the face width ratio. The facial features that defines the shape of the face are the face width, the face height, the bottom half of the face height and the chin to cheek average distance of both side. These facial features are calculated based on the nose tip, the nose root, the subnasale, the left and right nose alas, the left and right cheeks and the chin. The algorithm produce high accuracy of landmarks locating and features measuring. Over 70% of 3D facial models were successfully located all landmarks correctly. The facial features' measurements produced by the algorithm are guarantee over 85% of measuring accuracy.
The respirator prediction algorithm find the key facial features by comparing the prediction performance with different combination set of the facial features. The facial features that makes the prediction performance increase are the key facial features. The performance is compared by the precision of predicting the respirator is fitted. The algorithm found that each type of respirator has different set of key facial features. For 3M 1860s, the key facial feature is found to be the chin to cheek average distance of both side. For 3M 1862, the key facial feature is found to be the height of the bottom half face. For KCS, the key facial features are found to be the vertical angle of the nose and the nose width to the face width ratio. With the key facial features found in this project, the overall prediction performance is increased.
Copyright Tam Sze Wah Solomon and Andrew Lui 2014
Jonathan handles all external affairs include business development, patents write up and public relations. He is frequently interviewed by media and is considered a pioneer in 3D printing products.
After graduating from OUHK, Krutz obtained an M.Sc. in Engineering Management from CityU. He is now completing his second master degree, M.Sc. in Biomedical Engineering, at CUHK. Krutz has a wide range of working experience. He has been with Siemens, VTech, and PCCW.
Hugo Leung Wai-yin, who graduated from his four-year programme in 2015, won the Best Paper Award for his ‘intelligent pill-dispenser’ design at the Institute of Electrical and Electronics Engineering’s International Conference on Consumer Electronics – China 2015.
The pill-dispenser alerts patients via sound and LED flashes to pre-set dosage and time intervals. Unlike units currently on the market, Hugo’s design connects to any mobile phone globally. In explaining how it works, he said: ‘There are three layers in the portable pillbox. The lowest level is a controller with various devices which can be connected to mobile phones in remote locations. Patients are alerted by a sound alarm and flashes. Should they fail to follow their prescribed regime, data can be sent via SMS to relatives and friends for follow up.’ The pill-dispenser has four medicine slots, plus a back-up with a LED alert, topped by a 500ml water bottle. It took Hugo three months of research and coding to complete his design, but he feels it was worth all his time and effort.
Hugo’s public examination results were disappointing and he was at a loss about his future before enrolling at the OUHK, which he now realizes was a major turning point in his life. He is grateful for the OUHK’s learning environment, its industry links and the positive guidance and encouragement from his teachers. The University is now exploring the commercial potential of his design with a pharmaceutical company. He hopes that this will benefit the elderly and chronically ill, as well as the society at large.
Soon after completing his studies, Hugo joined an automation technology company as an assistant engineer. He is responsible for the design and development of automation devices. The target is to minimize human labor and increase the quality of products. He is developing products which are used in various sections, including healthcare, manufacturing and consumer electronics.
| Course Code | Title | Credits | |
|---|---|---|---|
| COMP S321F | Advanced Database and Data Warehousing | 5 | |
| COMP S333F | Advanced Programming and AI Algorithms | 5 | |
| COMP S351F | Software Project Management | 5 | |
| COMP S362F | Concurrent and Network Programming | 5 | |
| COMP S363F | Distributed Systems and Parallel Computing | 5 | |
| COMP S382F | Data Mining and Analytics | 5 | |
| COMP S390F | Creative Programming for Games | 5 | |
| COMP S492F | Machine Learning | 5 | |
| ELEC S305F | Computer Networking | 5 | |
| ELEC S348F | IOT Security | 5 | |
| ELEC S371F | Digital Forensics | 5 | |
| ELEC S431F | Blockchain Technologies | 5 | |
| ELEC S425F | Computer and Network Security | 5 |
| Course Code | Title | Credits | |
|---|---|---|---|
| ELEC S201F | Basic Electronics | 5 | |
| IT S290F | Human Computer Interaction & User Experience Design | 5 | |
| STAT S251F | Statistical Data Analysis | 5 |
| Course Code | Title | Credits | |
|---|---|---|---|
| COMPS333F | Advanced Programming and AI Algorithms | 5 | |
| COMPS362F | Concurrent and Network Programming | 5 | |
| COMPS363F | Distributed Systems and Parallel Computing | 5 | |
| COMPS380F | Web Applications: Design and Development | 5 | |
| COMPS381F | Server-side Technologies and Cloud Computing | 5 | |
| COMPS382F | Data Mining and Analytics | 5 | |
| COMPS390F | Creative Programming for Games | 5 | |
| COMPS413F | Application Design and Development for Mobile Devices | 5 | |
| COMPS492F | Machine Learning | 5 | |
| ELECS305F | Computer Networking | 5 | |
| ELECS363F | Advanced Computer Design | 5 | |
| ELECS425F | Computer and Network Security | 5 |
