Facilitate Personal and Community Connection

2021-2022 Capstone Design Projects

Project AI-51: Social Prescription App

Project Client: Beyond the Conversation

Purpose: Since the pandemic began, many people have been in self-isolation, experienced loneliness, and have become socially disconnected. Regardless of the effects of the pandemic, people are becoming more and more isolated despite technological advancements. Our app, Interlinked, aims to rebuild connections between people who have suffered from social loneliness and isolation by providing a platform to find groups and chat.

Major design contribution: For the application frontend, we used React and Typescript to construct a Model-View-Controller system allowing for easier mocking of the view and more efficient unit testing. In regards to the backend, we constructed a server using express and nodejs to handle all the API calls from the user browser and a remote MongoDB database that is used to store data necessary for our backend to run. Additionally, we have a recommendation engine to support resources and chat room recommendations. We chose collaborative filtering for the engine to improve recommendation accuracy based on user preference and interest.

Contact information: Jingyuan Wang: jingyuan981123@gmail.com; Kevin Yu: kaikailongwang3@gmail.com; Kelly Wong: kellywong48357@gmail.com; Henry Mao: qchenry.mao@gmail.com; Weihao Xia : rabbey@sina.com

Project CG-47: Open Source Audio Filters for Homeworld 3 and Hardspace: Shipbreaker

Project Client: Blackbird Interactive

Project Description: The majority of audio designers have no programming knowledge. This raises an issue, as audio plugins for video game engines are coded in languages such as C++. To solve this issue, audio designers use audio based visual programming languages such as Pure Data, in combination with audio software such as Wwise. In order for Pure Data to be used with Wwise, audio designers rely on tools such as the Heavy Compiler Collection (hvcc) to convert Pure Data files into plugins for Wwise. However, while Wwise has constantly been upgrading, the development of the hvcc tool has been discontinued, and no longer supports the latest versions (2021 and up) of Wwise. Fortunately, hvcc has been made open source. Our client, Blackbird Interactive, has requested that our team upgrade the hvcc tool to use the latest versions of Wwise, as well as add new features; adding support for Pure Data fast fourier transform (FFT), threshold, and Vline objects, as well as a Graphical User Interface (GUI).

In order to maintain version compatibility we needed to update the interface the plugins use to interact with the Wwise engine as well as create compilation options to allow for backwards compatibility. For creating the new object compatibility, we needed to interpret the Pure Data objects into representations the compiler could understand and recreate each of their functionalities in order to perform the same operations with the signals we were receiving from the audio engine in Wwise. The GUI was created with ease of use in mind such that anyone would be able to compile their audio plugins without needing to interact with a command line or move any files manually.

Contact information: Zi Tan: zitan39@gmail.com; Eleiah Hengeveld: EleiahHengeveld@gmail.com; Brandon Chan: brandonchan03@gmail.com; Abhinav Subramani: abhisubramani3@gmail.com; Tim Degerness: timdegerness@gmail.com

Project CG-48: Room-Based Procedural Game Level Generation

This team will not have a booth at the Design & Innovation Day due to confidentiality agreements.

Project Client: St Paul’s Hospital and UBC Emergency Medicine

Project Description: The main goal of this project is to develop an algorithm that can randomly generate indoor spaces for video game levels. The focus of the project will be to generate discrete, inter-connected rooms, such as office buildings or spaceship interiors. Our project consists of a node graph editor and an algorithm for generating and rendering a game level based on user inputs.

Project JY-15: Proximy Places

Project Client: Proximy Technologies Inc.

Project Description: With COVID-19 and increased social media consumption, people are feeling more isolated now than ever. The WHO reports that up to 1 in 3 adults globally are lonely, and that social isolation and loneliness can shorten lives, and damage physical and mental health. Although loneliness is a serious problem, people often struggle to break out of it. It is difficult for people to know how to begin discovering new connections and places around them based on the things they love.

Proximy Places addresses this growing concern by connecting its users with organizations based on their interests. Through a map-based feature, users can quickly and easily discover places that may interest them. Proximy Places also sends users periodic notifications about nearby places while the app is not in use, so that users never miss out on any opportunities around them.

Project PB-12: Immersio Language Learning Mobile Application

This team will not have a booth at the Design & Innovation Day due to confidentiality agreements.

Project Client: Immersio Learning Incorporated

Purpose of project: To help save endangered languages by creating an intuitive language learning platform for teachers.

Major design contribution cannot be discussed to to NDA/IP.

Project PB-78: Smart Restaurant Analytics Dashboard

This team will not have a booth at the Design & Innovation Day due to confidentiality agreements.

Project Client: DYNE Technologies Inc.

Project Description: The Smart Analytics System is designed to categorize and determine the sentiment of restaurant reviews. Business owners can quickly learn what their customers think about the food, price, service, and ambience without having to manually sift through hundreds or thousands of reviews.

Project PL-01: Skyline innovation is using ML and NLP to build a next generation SaaS video chat app

Project Client: Skyline Innovation

Project Description: The purpose of our project is to build a machine learning model for word correction as well as a web application to host the model in order to interface with users. The model and web app together serve as a proof of concept for our client’s ultimate goal, an end-to-end AI video chat app designed for augmenting the conversation where mispronounced words interrupt the flow of a conversation. As the user speaks into the microphone, the app provides a real-time transcription feed and identifies mispronounced words, thus providing active feedback to the speaker. This will improve the lives of millions of people with speech impairments, especially those who are suffering from Parkinson’s disease, and people they communicate with.

To obtain transcripts of conversations we have investigated and profiled existing automatic speech recognition systems from various companies and integrated a speech-to-text module into our application. We have also come up with mechanisms to generate features for each word input to facilitate word error detection and correction in the machine learning model that we then connect with an integrated speech-to-text module. Due to a lack of freely available audio samples from people with Parkinson, we have developed several heuristics for generating data that approximates transcripts of impaired speech. We used these heuristics of word replacements on various corpuses of English conversations to generate sets of data for model training. The model is then trained on above samples to tune parameters, the weights of the connections between nodes of the network, to find the optimal match to perform word correction.

Finally, we have also constructed a set of metrics for validating how well our model has been trained to perform word correction tasks. They can be expressed in terms of accuracy, true positive rate (TPR) and false positive rate (FPR)

Team members contact info: Alek Dudko: 604-760-8229, alekdudko@protonmail.com; Grady Thompson: 778-957-3702, grady_thompson@outlook.com; Yuefeng Zhao (Max): 236-990-2831, maxonzhao@gmail.com; Conrad Cheung: 778-883-1398, conradftw@hotmail.ca; Jordan Schneider: jordanschneider@telus.net, schneiderjord@aquinas.org

Project PN-86: Instagram Smart Insights

Project Client: BroadbandTV Corp

Project Description: Hashtags play an important role for content discovery on social media platforms like Instagram. Content creators can attract more followers by using relevant and popular hashtags in their media posts.

The goal of this project is to build a hashtag recommendation and ranking system to help Instagram content creators reach a wider audience, using hashtags generated by various pre-trained image-based and text-based deep learning models. We provide unique hashtag rankings for each content creator using content-based filtering. The other contributions of the team include a new dataset collected for potential future training and validation and a user interface that allows users to easily upload their posts, login to their Facebook account, and get personalized hashtag rankings.

Project SF-40: Emerging Media Lab CAVE: Virtual Reality in a Box

Project Client: UBC Emerging Media Lab

Project Description:

Virtual reality (VR) can be inaccessible to some users who cannot easily use or obtain a VR headset.

GRoTTO is a alternative virtual reality system that projects a virtual environment onto a set of screens around a user. Using a depth camera and body-tracking software, the user can then move within the environment and have the images move with them. GRoTTO can be implemented with easily-accessible resources and is flexible in configuration, and is perfect for both personal and organizational use.

Contact Information: emergingmedia.lab@ubc.ca

Project SF-43: IoT sensor to improve learning and focus in classrooms

Project Client: Airtame ApS

Project Description: “IoT sensor to improve learning and focus in classrooms” is the project that our team (SF-043) has undertaken. The purpose of this project was to design a solution that will aid educators in making decisions and adjustments based on environmental classroom conditions. Studies have shown that classroom environmental conditions play a key role in optimizing student learning.

Major Design Contribution: This project aims to provide actionable recommendations and meaningful representation of the conditions within a classroom setting through the use of IoT sensor hardware. The sensing hardware that our group has designed is focused on sensing and reporting the temperature, humidity, and CO2 environmental conditions within a classroom setting.

Contact Information: Matt Fournier: https://www.linkedin.com/in/matt-fournier/ ; Mitchell Gordon: https://www.linkedin.com/in/mitchell-gordon0/ ; Larry Qian: https://www.linkedin.com/in/larry-q-4786a2230/ larryqian@alumni.ubc.ca ; Lam Hoang: https://www.linkedin.com/in/tunglam-hoang/ ; Valentine Ssebuyungo

Project TL-41: Novel ML solutions to quantify motor performance in people with Parkinson’s disease

Project Client: Pacific Parkinson’s Research Center

Project Description: Our client, the Pacific Parkinson’s Research Centre has collaborated with their partners at NTU in Singapore to create an app to automate the collection of data on Parkinson’s Disease symptoms. Our team has taken raw data from test subjects who have used the app and created machine learning models to predict the severity of Parkinson’s Disease without drawn out tests.