Learning from Data: Pattern Recognition and Classification, Machine Learning, Data Analysis/Mining, Discriminative Training and Error Minimization, Statistical Signal Processing, Brain-Computer Interface Design, Speech, Speaker, and Spoken Language Recognition, Protein Folding, Motion Capture and Analysis

Medical Embedded Systems: Algorithm Design, Collaborative Signal and Information Processing, Power-Aware Computing, Design and Analysis of Wearable Computers, Real-time and Self-Calibrating Devices, Cyber Physical System Design, Scalable Learning, Context-Aware Reasoning

Wireless Health & Biomedical Applications: Body Sensor Networks, In-Home Rehabilitation, Physical Movement Monitoring, Diagnosis of Neurodegenerative Disorders, Biofeedback for Movement Disorders, Sports Training

Wearable Health Monitoring

Wearable health technology is drawing significant attention for good reasons. The pervasive nature of such systems providing ubiquitous access to information will transform the way people interact with each other and their environment. The resulting information extracted from these systems will enable emerging applications in healthcare, wellness, emergency response, fitness monitoring, elderly care support, long-term preventive chronic care, assistive care, smart environments, sports, gaming, and entertainment which create many new research opportunities and transform researches from various disciplines. Despite the ground-breaking potentials, there are a number of interesting challenges in order to design and develop wearable medical embedded systems. Due to limited available resources in wearable processing architectures, power-efficiency is demanded to allow unobtrusive and long-term operation of the hardware. Also, the data-intensive nature of continuous health monitoring requires efficient signal processing and data analytics algorithms for real-time, scalable, reliable, accurate, and secure extraction of relevant information from an overwhelmingly large amount of data. Therefore, extensive research in their design, development, and assessment is necessary.


Embedded Processing Platform Design

The majority of my work concentrates on designing wearable embedded processing platforms in order to shift the conventional paradigms from hospital-centric healthcare with episodic and reactive focus on diseases to patient-centric and home-based healthcare as an alternative segment which demands outstanding specialized design in terms of hardware design, software development, signal processing and uncertainty reduction, data analysis, predictive modeling and information extraction. The objective is to reduce the costs and improve the effectiveness of healthcare by proactive early monitoring, diagnosis, and treatment of diseases (i.e. preventive) as shown in Figure 1.

Figure 1. Embedded processing platform in healthcare




Multimodal Driver Monitoring Platform Design

The objective of this research is to investigate a novel proactive driver monitoring safety platform design via heterogeneous wearable and in-vehicle sensor networks. This effort is a potential advancement in the next generation motor vehicle technology that would incorporate information about driver biological and behavioral state to facilitate driving experience. We develop a reliable model of the associations between multiple indicators of biological and behavioral state and the act of driving as well as changing conditions on the road to detect and predict high risk conditions and provide the driver with feedback. The research plan aims to advance the field of vehicle research by developing methods for: (1) Collecting biometric data during actual driving on the Michigan Test Facility using our proposed integrated wearable driver monitoring platform consisting of a network of multiple heterogeneous body worn sensor network, in vehicle sensors, and mobile data; (2) Optimize different modules of our proposed platform by performing sensor selection and localization, data flow enhancement, resource optimization, and efficient bio-feedback design; (3) Analyzing the data in order to develop an accurate estimate of the driver model that integrates his/her biological state and driving behavior as well as environmental factors.



We are investigating the development of highly adaptive, scalable and resilient driver monitoring systems that integrate heterogeneous wearable, in-car, and environmental sensors for monitoring and intervention. This investigation intends to extend and facilitate current infrastructure and vehicle technology to include real-time driver monitoring and feedback in order to integrate them into the information shared between the driver and vehicle (D2V) or driver and infrastructure (D2I).


EEG-Based Driver Distraction Analysis and Detection

Driver distraction is a significant cause of accidents leading to injuries and fatalities on the roadway. Driving is a complex task and demands continuous visual and cognitive attention on the primary task of driving. Drivers are at risk of responding more slowly or less suitable to intricate and dynamically changing situations that entail their complete attention. Therefore, monitoring and detection of driver distraction can directly facilitate decreasing the costs associated with roadway disasters. There has been much research effort toward developing automatic detectors of drivers’ distraction state. Many approaches were based on analyzing the driving behavior or using camera-based techniques to monitor and evaluate driver distraction. However, besides the issues with privacy and consumer adoption, there are limitations with how early they can detect the signs of distraction. Neurophysiological signals such as Electroencephalogram (EEG) and brain activity analysis have been extensively used to understand the precursors of distraction at the physiological level to develop systems to alert the drivers well in advance. The objective of this study is to take a step toward establishing a systematic framework to extract effective descriptors and to measure the impact of in-vehicle secondary tasks on driver cognitive state during naturalistic driving by capturing the changes in EEG dynamics. We employed our wearable data acquisition platform to collect wireless EEG data from six subjects during a naturalistic driving session and investigated six potentially distracting stimuli. We present a standard analysis framework to examine the impact of various EEG signal pre-processing, feature extraction, and classification methods in order to detect driver engagement in a secondary task with high accuracy solely based on their recorded EEG.

Wearable Brain-Computer Interface

Steady State Visual Evoked Potential (SSVEP) has been commonly adopted in Brain Computer Interface (BCI) applications. For wearable BCI applications, several aspects of SSVEP-based BCI systems, such as speed, subject variability, and accurate target detection are under ongoing research investigations. Up to date, Canonical Correlation Analysis (CCA) has been considered state-of-the-art feature extraction method for SSVEP-based BCI systems. Nevertheless, although CCA outperforms traditional SSVEP detection methods, such as Power Spectral Density Analysis (PSDA), achieving high accuracies when detecting target frequencies is still a challenging task due to user variation and physiological changes in the human body. In our latest effort, we investigate an SSVEP-based BCI application using wireless EEG recording and Android tablet-based user interface. We propose a fusion of CCA and PSDA solutions at the score level by dividing their score space into multiple partitions, and extract and combine their complementary discriminative information to minimize the detection error. We investigated transforming the fusion score space to lower dimensions with the purpose of alleviating the redundancy. For this purpose, we employed Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Generalized Discriminant Analysis (GDA). Our experimental results demonstrated that our proposed score fusion method combined with a kernel approach to dimensionality reduction is effective in reducing the effect of noise and non-stationary elements in EEG dynamics and achieved significantly higher detection accuracies.

Characterization of Driver Distraction During Naturalistic Driving Using ECG Signals

Many of the fatalities involved in accidents on the road are associated with driver distraction. In order to reduce the possible chances of road disasters, it is essential to characterize the pre-requisites of driver distraction. While driving, the driver might get distracted in several ways such as talking on the cell phone, texting, and having a conversation with the passenger. There has been extensive research conducted to estimate driver state in recent years particularly using camera-based systems. However, camera-based systems face challenges such as privacy or latency in detection. Using physiological signals to identify distraction such as Electroencephalography (EEG) has been shown to accomplish more reliable detection. However, EEG-based detection systems require intrusive implementation and complex signal processing. On the other hand, Electrocardiogram (ECG) is a reliable signal which can characterize the physiological changes consistently, with minimal intrusiveness, and at low cost.  In this paper, we propose an ECG signal processing recipe with the aim of predicting driver distraction in real-time. Eight drivers actively participated in the naturalistic driving experiment where distraction was induced by: 1) making a phone call and 2)  having an active conversation between the driver and the passenger. We present an effective frequency subBand analysis using Wavelet Packet Transform (WPT) to localize the impact of distracting elements. Due to high dimensionality of the original WPT features, we then applied Linear Discriminant Analysis (LDA) for feature space dimensionality reduction; preserving discriminative capability of the predictive model. In order to further enhance the prediction ability of the system, we used Kernel transformation in order to take into account non-linear interactions of the input feature space.

Gait Pattern Recognition and Authentication Using Wearable Inertial Sensors

Motion ability is one of the most important human properties, including gait as a basis of human transitional movement. Gait, as a biometric for recognizing human identities, can be non-intrusively captured signals using wearable or portable smart devices. In my study gait patterns is collected using a wireless platform of 5 sensors located on the body of a population of subjects and they have been asked to walk in different conditions: various slope, speed, and environments. Using time-frequency analysis and deep learning, we have been able to detect the right subject with ~95% accuracy. We are extending the work to a pervasive and ubiquitous setting and adaptively learn from subjects motions.

Time Series High Dimensional Representation and Genetic Algorithm Feature Selection

Feature generation and feature selection are two of the most important aspects of predictive modeling. In many cases the use of good features is more beneficial to prediction power than the choice of algorithm itself. This project uses a time series feature generator and correlation filtering plus genetic algorithm for feature selection. The overall goal is to be able to predict workload changes based on various biometric signals.

Determining the treatment quality of Scoliosis in Adolescents: A brace monitoring system

Scoliosis is a medical condition, which often occurs in adolescents, and causes an individual’s spine to have an irregular curvature. Monitoring the effectiveness of treatment of Scoliosis is a challenge for physicians.   To monitor the patient’s treatment, a low power multi-modal sensor board, is embedded onto the brace of the patient. The hardware consists of a sensor board, a pressure sensor, an accelerometer and a gyroscope. We collect the patient data and analyse it to evaluate the effectiveness of treatment.