Imran Ture

Deep Emotion Recognition using Wearable Sensors

Emotion recognition is an emerging interdisciplinary field that integrates methodologies from affective computing, sentiment analysis, signal processing, and machine learning. This project focuses on classifying emotions such as amusement and stress using physiological signals like heart rate and skin conductivity, collected via wearables. To build the model, a hybrid approach was employed that combines various deep learning architectures, including Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. This approach is capable of accurately capturing changes in physiological signals to identify specific emotional states.

The performance of the model has been rigorously evaluated using a variety of metrics, including accuracy, precision, recall, and F1-score. To fine-tune the model's performance, hyperparameter tuning was conducted using grid search and Bayesian optimization techniques. The model achieved an accuracy rate of 92% under a 5-fold cross-validation setting. As wearable technology continues to evolve, this project serves as a significant contribution to the development of real-time affective computing systems that could be integrated into future generations of a wide range of wearable products, from smartwatches to health monitors.

E4 TimeStamper: A GUI application for automatic timestamping and analysis of physiological signals collected from Empatica E4 wristbands

E4 TimeStamper is a user-friendly GUI application designed to facilitate researchers in adding timestamps to physiological signal data obtained from Empatica E4 wristbands. This tool allows for seamless extraction and precise timestamping of files, all in accordance with the chosen timezone and preferred date & time format.

Image source

E4 TimeStamper, adopted and implemented by researchers worldwide in their studies,¹ is available for both Windows and Mac operating systems. For more information on how to download and use the app, please go to the GitHub page.

View on GitHub

Python Implementation of OpenIntro Statistics Labs

We developed the Python labs for OpenIntro Statistics, an open-source textbook for introductory statistics used at many universities (Community Colleges to the Ivy League) around the world, to promote the understanding and application of statistical data analysis using Python.

The labs are officially available on the OpenIntro Statistics website.

BBL-PhysioDB: Data Hub for Entrepreneurship Assessment and Development Workshop

BBL-PhysioDB is a robust database, developed with Django and PostgreSQL, for the Entrepreneurship Assessment and Development Workshop studies conducted at RMIT's Behavioural Business Lab (BBL). The database can effectively manage data collected from over 150 participants, including entrepreneurs, artists, and professionals with diverse backgrounds. It has the capability to store extensive data collected through multiple lab experiment sessions, including physiological signal data recorded through E4 wristbands from Empatica, along with 200+ questionnaire responses providing demographic information and insights into participants' entrepreneurial engagement.

Image source

The database offers a user-friendly interface that simplifies navigation and enables robust search capabilities. Researchers can seamlessly query the database to explore findings from BBL studies. The database scheme can be leveraged by other researchers to address their own research needs.

Advancing Water Safety: An Early Warning System to Monitor and Evaluate Drinking Water Quality

An early warning system was developed that can enhance the safety and quality of drinking water in Turkiye by identifying hazardous contaminants such as E. coli O157:H7, anthrax, and ricin. Utilizing a hybrid approach that integrates various statistical methods like Z-score analysis, moving averages, control charts, and weighted voting, the system was designed to provide real-time monitoring and detect unexpected levels of multiple parameters, including temperature, pH, total organic carbon, conductivity, oxidation-reduction potential, free chlorine, and dissolved oxygen.

The idea with employing a multi-tiered approach is to facilitate early intervention for minor deviations while enabling immediate action for more severe anomalies. It has the potential for wide-ranging applications, not just limited to municipal water treatment plants but also extending to main storage and distribution lines, thereby serving as a critical tool with the capability to safeguard public health and environmental integrity.