Engineering Projects

Developing an Upper Limb Rehabilitation Robot for Robot-Aided Therapy (2024): 

The recovery of dis-coordination in arm movements hinges on activity-induced alterations in the brain, yet existing rehabilitation approaches are labor-intensive and heavily reliant on therapists. There's a pressing need for robotics to facilitate this process, enabling repetitive movements and alleviating the burden on therapists. This project sought to develop a rehabilitation robotic device capable of guiding upper limb motion within a 2-D plane while collecting data on arm movements. It features adjustable resistance levels to accommodate various exercise intensities while ensuring adjusted resistance during arm movement.

Detailed Information about this project can be found here: 

Asymmetric gait detection in real-time using recurrent neural network (2023): 

The objective is to employ AI techniques in both the capture and analysis stages of gait asymmetry, facilitating early detection of abnormal characteristics to administer timely treatment. The AI methodologies employed include computer vision for human pose estimation - a machine-learning application that identifies human postures and positioning, and Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN) - another machine learning technique renowned for examining temporally based data. 

The entire process involves participants taking part in asymmetric walking trials (on a split-belt treadmill), recorded using human pose estimation and a singular webcam. Subsequently, the data from this trial is extracted and evaluated by the RNN for real-time evaluation. 

Detailed Information about this project can be found here: 

EMG Controlled Racing Car (2023): 

This project uses an Arduino that controls two DACs (digital to analog converters) that output certain voltages to the remote control of a remote-controlled car. The voltages outputted by the DACs act as the remote's potentiometers causing the car to drive or steer. EMG amplifiers with electrodes placed on the arm send EMG signals to the Arduino. The Arduino runs a smoothing algorithm on the data and then outputs that data to the DACs, which control the car, resulting in a toy car controlled by EMG signals.

Detailed Information about this project can be found here: 

Video Game Rehabilitation (2023): 

As machine learning tools become increasingly prevalent in the medical industry, it is essential to leverage the advantages of such emerging technology to enhance rehabilitation programs. The present demonstration is a Python program developed using MediaPipe (a Google-authored AI library that enables human pose estimation) and other computer vision capabilities like OpenCV. The outcome is a simple video game where the user can interact with circles on the screen, aiming to hit a moving circle within a given time limit, using either their hands or feet. Tools like this hold promise for future applications in rehabilitation programs, fostering motor learning through actively engaged movement training.

Detailed Information about this project can be found here: 

Motion Capturing-based Robotic Arm Control (2021): 

In this project, a participants’ arm kinematics (angles) are measured using the motion capture markers (PhaseSpace system) attached on their arms, and the kinematics data are used to compute the joint angles to control the robotic arm (Reactorx 150) 

Python programming was used to measure marker position data, compute Euler angles (shoulder angle, elbow angle, wrist tilting angle, and wrist rotation), and control the robotic arm. 

Applications of an EMG Driven Bionic Hand Using Deep Learning Algorithm (2019):

This project explored and implemented one of AI algorithms (deep learning algorithm) that can classify different hand/finger gestures and developed a muscle-controlled robotic arm mimicking the user's hand gestures. When a user presented different hand gestures, muscle signals were acquired through surface sensors attached to his/her forearm and the developed control system successfully recognized the gesture patterns based on the deep learning algorithm (artificial neural network) implemented. This project demonstrated the feasibility of the artificial neural network that could classify 8 different finger gestures using multiple EMG channels, which were placed on the forearm.

Internet of Things: Door sign application (2018):

The Internet of Things is a global network of computers, sensors, and actuators connected through internet protocols. A most basic example is a PC or a smartphone that communicates over the Internet with a small device, where the device has a sensor or an actuator attached (e.g., a temperature or a electric motor). Here is an example where a door sign is controlled from anywhere using a web browser.

>> How to start developing an IOT applications

Applications of an EMG Driven Bionic Hand Using Principal Component Analysis (2016):

The final goal of this project was to develop an EMG control algorithm that would give subject's total control. To test the control system, 8 EMG sensors (Tringo, Delsys) were placed on the arm, and the EMG signals were recorded while the user produced 7 different postures. The control program, implemented by the PCA algorithm, was able to process the EMG signals and correctly recognize the users' 7 different postures. 

Muscle Cars (2013):

The EMG Race Car was designed to allow a user to power a device (i.e. the race car) using their muscles. Electrodes were attached to the user's muscles to gather data from muscle activity. This data was then converted into readable data for the race car to determine its speed. >>Need more information about EMG-control? 

>>Read "Technical Document for EMG-control"