Mengxuan Ma

Ph.D

Data Science

University of Missouri-Columbia

Graduate Research Assistant in the Center

- For Eldercare and Rehabilitation Technology, College of Electrical and Computer Engineering - Daily Activity Recognition and Assessment System for Stroke Rehabilitation - Worked in an interdisciplinary team to design a daily activity recognition and assessment system for evaluating the quality of daily motions of stroke patients - Developed an activity logging system to record depth frames and skeletal joint data of a stroke patient. Investigated - Proposed a convolutional neural network-based ensemble model to detect actions from the depth videos - Tested the algorithm on real-life cooking videos in four different kitchen environments with ten participants. The - Performed kinematics assessments on recognized actions. The assessment results of patients' daily motions help - Movie Recommendation System - Implemented a movie recommendation system using Collaborative Filtering algorithms - To predict the ratings for movies accurately, two models were investigated. An Alternating Least Squares (ALS - Implemented models were evaluated using the MovieLens 1M dataset. 70% of the data was randomly selected for - Long-term Kinect-based Stroke Rehabilitation Game Assessment - Implemented a motion assessment module in a Kinect-based stroke rehabilitation game to quantitatively evaluate the upper-body movement of stroke patients - Preprocessed the collected data by removing the duplicate samples, filling the missing samples and filtering the noise - Performed kinematics assessment using range of motion measures, efficiency and smoothness measures. Applied - Evaluated the structure of hand movement trajectories by applying a density-based algorithm, OPTICS - Conducted Welch's ANOVA and post hoc tests which showed that there were significantly different among the assessment outcomes of healthy, affected and unaffected sides on the data of 8 stroke players and 30 healthy players - Angel-Echo: A Personalized Health Care Application - Designed a personal health monitoring system using an Angel sensor and an Amazon Echo that allows a user to acquire health information with a speech interface - Developed a system that received health information (heart rate, steps, and skin temperature) collected from an Angel - Sensor via Bluetooth GATT Protocol, and then stored the data to the Amazon DynamoDB database - Designed an Amazon Echo skill on Amazon Web Service (AWS) to handle users' health-related requests, fetch the data from the database, organize the data to speech responses using Alexa Voice Service - Tested the Amazon Echo speech recognition accuracy on different populations. The misunderstanding rate from the young adults was 2.6% lower than the rate from elderly adults