HUI Zheng

Hi, I'm Hui Zheng, a PhD (HCI field, with focus on Assistive technology for people with cognitive/mental impairements ) grauated from George Mason University graduated at May 2021. Then, I worked as a software engr for a digital therapeutics project for people with Schizophrenia at Click Therapeutics (a startup focuding on Digital Therapeutics for mental health) till the August 2022. And I started working at Google in the KE feedback team.

My research interests are Human-computer interaction, especially applying Machine Learning in assistive wearable technologies for people with cognitive/mental impairements or with special needs. I am interested in applying AI in web, mobile, or wearable assitive applications for people with Neurodiversity. I have publications at top HCI conferences including CHI, ASSETS, InterAct, and etc. And I am in the review comunity for CHI.

Besides, I also have experince in Machine (Deep) Learning and Computer Vision. For internship experiences, I have interned at Microsoft Research (HCI ressearch) and Intel Labs (Computer Vision research).

Publications and Patents

[ Google Scholar Link ]

WLA4ND: a Wearable Dataset of Learning Activities for Young Adults with Neurodiversity to Provide Support in Education. ASSETS 2021.
Lessons Learned in Designing AI for Autistic Adults. ASSETS 2020.
Evaluating WELI: A Wrist-Worn Application to Assist Young Adults with Neurodevelopmental Disorders in Inclusive Classes. InterAct 2019.
Assisting Students with Intellectual and Developmental Disabilities in Inclusive Education with Smartwatches. CHI 2018.
Designing a wearable technology intervention to support young adults with intellectual and developmental disabilities in inclusive postsecondary academic environments. Journal of Special Education Technology 2018.
WeLi: A Smartwatch Application to Assist Students with Intellectual and Developmental Disabilities. ASSETS 2017.
Wearable Life: A Wrist-Worn Application to Assist Students in Special Education. HCII 2017.
Label Propagation in RGB-D Video. IROS 2017.
GEST-DC: Unifying Transportation and Air Quality Information in an mHealth Application. AHFE 2019.
Patent: cn zl 2014 1 0366737.5 Video Salient Object Segmentation Based on Superpixels. With Chinese Academy of Sciences.

Selected work

Selected Key senor-specific features Visualization of reading activity generated via the attention model for four participants . P2(a) and P7(b) are with Down Syndrome. P3(c) and P6(d) are with ADHD.

WLA4NDD - inclusive wearable sensor dataset [paper link, dataset link ]

WLA4NDD, a dataset collected from smartwatch built-in sensors (accelerometer, gyro, and heart rate) collected hand gestures from eight young adults with neurodevelopmental disorders (NDD) performing learning tasks. The dataset provides a novel benchmark to train activity recognition models for users with NDD. It aims at enhancing the assistive technologies using machine learning to support students with NDD in inclusive education. A classification using state-of-the-art models ( both deep and conventional models) has obtained accuracy of 97.1% (user-independent). This work has been accepted at ASSETS 2021 (The top conference in the Accessibility .field)

Wearable Life [ paper link 1, paper link 2 ]

Wearable Life, a wearable and mobile app for young adults with intellectual and developmental disabilities(IIDs) to help with the coordination and communication between the students and their in-class assistants in inclusive education in a less obtrusive way, which provided timely intervention notifications sent automatically or manually by assistant to student, including the features of interventions and self-mood assessment, including 14 user studies (interviews, focus groups, surveys, usability tests, field studies). The related work has been accepted at CHI 2018 (The top conference in the HCI field), ASSETS 2017, and etc..

Human-AI sci-fi collaborative writing [Website Link (Chinese)]

A web application for Human-in-loop AI science fiction (sci-fi) collaborative writing, which enables users to write sci-fi with next sentence/plot suggestions generated from a GTP-2 based NLP model, according to the context. The user can pick the generated sentences, edit the them, or write own words. Facilitated writing speed to 500 ~ 1000 words/minute. Won an AI Innovation Application Champion.

VC4A-AI video calling for autistic adults [project link]

VC4A is a video calling website for people with autism with real-time emotion feedbacks during video calling. The visualized emotion feedbacks on web video screen from two sources: 1). a Wizard of Oz website to control emotion metrics manually, 2). softMax results from an existing affective computing machine learning model. I was maily focused on the development part when I interned at Microsoft research. This work has been accepted at ASSETS 2020.

Label Propagation in RGB-D Video [paper link]

This aim is to propagate semantic labels in RGB-D video of indoor scenes given a set of ground truth keyframes. It uses the camera poses and 3D point clouds for propagating the labels in superpixels computed on the unannotated frames of the sequence. The propagation task is formulated as an energy minimization problem in a Conditional Random Field (CRF). Experiments on SUN3D dataset showed superior performance(increased IoU by 7.5%) to an optical flow based label propagation approach. This work has been published in IROS 2017 (one for top conference for robotics).

Publications and Patents

Selected work

Get in touch me at [ hfranceszheng@gmail.com, hzheng5@gmu.edu ]