School of Science and Technology 科技學院
Computing Programmes 電腦學系
| Programme | Bachelor of Computing with Honours in Internet Technology Bachelor of Science with Honours in Computer Science |
| Supervisor | Dr. Keith Lee |
| Areas | Virtual Reality Applications |
| Year of Completion | 2024 |
This report presents a VR prototype called Job Interview Simulator as a solution to reinforce better performance in first job interviews taken by fresh graduates.
The project aims to support fresh graduates in improving their job interview performance through a VR-based application. The key goals include:
The system is designed to provide feedback and track user progress, enhancing both language skills and interview readiness.
Demonstration Video
Presentation Video
Our solution leveraged existing hardware but innovated in software aspects, including backend scripts and user-facing programs.
The project has entered its final development phase using PICO 4 Pro VR headsets. Key components include:
VR Application Development:
Server-Side Processing:
Sample Q&A Module:
Chatbot Interviewer:
Voice Interaction:
Use Case Diagram and Function List:
There is only one user type in our application which is the user themselves. The user would first choose a VR scene and a question type before starting their job interview practice. They would be able to practice their interview responses in three different ways including picking a choice in multiple-choice questions, reading aloud a suggested answer, and answering questions in their own words. The details are shown as follows
Figure 1: Use Case Diagram and Function List
Figure 2: Application Hierarchy
Figure 3: Design Chart of Sample Q&A/Pronunciation Training
Once a VR user opens the Q&A module to start learning, Unity can visit the Python server by accessing a URL link. The Python server will be acknowledged to call one of its functions to collect all data stored in a JSON file. As mentioned before, the JSON file is deployed to store and update all questions, answers and explanations. The Python server then relays the data back to Unity so that VR user can see the questions, answers and explanations one by one.
Figure 4: JSON file for the Q&A Reading module
The system enables real-time pronunciation feedback in the VR interview app using a multi-layered architecture:
This setup combines cloud storage, real-time processing, and speech analytics to deliver detailed and automated feedback on spoken performance.
Figure 5: Design Chart of Speech Scoring System
3D models
Figure 6: Office Meeting Room Scene (Unity 3D)
Figure 7: Office Reception Area Scene (Unity 3D)
Question and Answer Reading
This module helps users improve speaking and presentation skills by allowing them to practice reading aloud sample job interview questions and answers, followed by detailed speech performance analysis.
Key Features:
Language Options: Users can choose to practice in English, Cantonese, or Mandarin, with the system adjusting accordingly.
Figure 8
Interactive Q&A: Users can view sample questions, answers, and explanations on a virtual panel and navigate through different Q&A sets.
Figure 9
Users record their spoken responses using a VR interface.
Figure 10
A color-coded word accuracy list helps identify specific pronunciation errors (green = excellent, red = poor).
Figure 11
Module- Virtual Interview
The module blends interactivity, real-time feedback, and multilingual support to enhance users' job interview readiness and spoken language proficiency
Key Features:
Start Interface: Users begin the session by clicking “Start” to enter the interview scene.
Figure 12
Language Options: Users can choose between English, Cantonese, or Mandarin before proceeding.
Figure 13
A microphone button appears after language selection.
Users record their responses by clicking the mic; input is disabled while speaking.
Figure 14
After recording, users can view their spoken text, delete it, or resend it.
Figure 15
Clicking “Send” transmits the response via ZMQ to a backend server.
Figure 16
The POE chatbot processes user input and returns a reply.
Figure 17
The response is converted to speech for playback.
Users can click “Replay” to hear the interviewer's response again.
Figure 18
If users do not look at the interviewer, “Look at the interviewer!” will appear to remind users.
Figure 19
Key Achievements
Limitations
