1: Design of an eNose for low-cost, non-invasive, non-irradiating and specific breast cancer screening

Proposed by:
Judit Giró Benet
jgirbene@uci.edu)

Faculty Advisor:
Prof. Fadi Kurdahi
kurdahi@uci.edu)

Time Size: 1 – 2

Description:

Whilst the number of diagnosed breast cancer patients is on the raise, the need for a low-cost, noninvasive and non-irradiating breast cancer screening device becomes more urgent. As a response to the current debate on the worthiness of the mammogram, a new opportunity has appeared for a novel breast cancer screening solution that performs with enough sensitivity and improved specificity with respect to that of the mammogram. The proposed project is a continuation of a previous bachelor thesis [link] that proved the following hypothesis:

“The study of the volatile organic compounds present in urine is sufficiently significant to enable class prediction among control subjects and metastatic breast cancer patients”

2: Autonomous Robot Navigation with Human Interaction

Proposed by:
Prof. Kwei-Jay Lin
klin@uci.edu

Description:
Program an indoor robot to travel from one position to another using existing indoor localization technology. The robot must avoid running into objects on its path. To identify routing path (or strategy), the robot needs to process camera images and sensor readings in its controller to calculate the current position and status in real time. It should also recognize human and identify human gestures to follow human instructions.

3: Attention or in Tension! – A brain computer interface application to detect mental attention states of a person.

Proposed by:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2 – 3

Description:
A wearable sensing system collecting brain signal through EMOTIV EPOC+ 14 channel mobile EEG. The device will be able to track the brain signal reflecting various human mental state in different situations (e.g. attentive/inattentive/drowsy) and generate notifications accordingly to meet user requirements (e.g. May be useful for ADHD patients).

4: Brainiac – A brain computer interface application to give commands using brain signals.

Proposed by:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2 – 3

Description:
A wearable sensing system collecting brain signal through EMOTIV EPOC+ 14 channel mobile EEG. The device will be able to track the brain signal reflecting various human moods in different situations (e.g. feeling dark) and generate commands accordingly to meet user requirements (e.g. turn on the lights).

5: DriveSafe – A brain computer interface application to monitor driver drowsiness using EEG signals.

Proposed by:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2 – 3

Description:
A wearable sensing system collecting brain signal through EMOTIV EPOC+ 14 channel mobile EEG. The device will be able to track the brain signal reflecting drivers’ drowsiness in different situations (e.g. drowsy/alert) and generate notifications accordingly to alert them (e.g. May ring an alarm in the mobile).

6: Well & Calm – A wearable device application to avoid stress and remain calm.

Proposed by:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2 – 3

Description:
A wearable wrist worn biosensors collecting electrodermal activity (EDA), temperature, acceleration, heart rate (HR), and arterial oxygen level (SpO2). The device will be able to track the physiological signals reflecting various neurological status (including physical stress, cognitive stress, emotional stress and relaxation) and generate notifications accordingly to help user to be calm (e.g. play soft music or suggest breathing).

7: LiDAR SLAM mapping system for autonomous driving system

Proposed by:
Wenhui Wang
wenhuiw3@uci.edu

Faculty Advisor:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2

Description:
The goal of this project is to build a mapping system based on LiDAR to perform 2D mapping (and 3D mapping as optional goal) based on SLAM algorithm. The project might be able to base on the autonomous car with LiDAR and NVIDIA GPU in Prof. Faruque’s lab. The system will scan the obstacle with LiDAR and generate the map of the environment in realtime. The ROS will be utilized in the project and a camera might also be involved. The vehicle should be able to run in any environment and sketch at least the 2D map in realtime basis.

8: Smart Driver Monitoring and Feedback using Smartphone

Proposed by:
Prof. Kwei-Jay Lin
klin@uci.edu

Description:
The project is to design an intelligent agent intended for improving automobile drivers’ performance by applying persuasive technology. Using the sensing capabilities of today’s smartphones, the project is to build the software for car motion detection, detect driving events, analyze event patterns for driving behavior classification, and finally produce driver feedback to improve driving performance. While many current driver management systems look into driving behavior as a single event (e.g., lane-changing), a driving behavior may be judged. based on the historical data. The project is to build an effective driver persuasive system. The first component is the personality classification, which recognizes drivers’ personalities by analyzing driving behavior patterns. The second component is the feedback generation, which determines the current driving behavior’s risk based on immediate behaviors. Once the system has identified the drivers’ personalities and risk of driving behavior, it should apply means of persuasive technology, friendly-feedback, and suggestion services to help drivers improve their behaviors.

9: Video Compression using Variational Auto-encoders

Proposed by:
Samruddhi Kahu
skahu@uci.edu

Team Size: 1 – 2

Description:
This project aims to improve compression performance using Variational Auto-Encoders. Recently, use of machine learning approaches to image compression has led to considerable performance improvements over classical image compression techniques. However, learning based video compression has started to be explored very recently. This area is in nascent stages of development and has a wide scope of improvement. According to Cisco, more than 82% of the internet traffic is estimated to be videos by 2022. Therefore, improved video compression would definitely have great economic impact. The project requires the study of variational auto-encoders, image and video processing and knowledge of the classical and learning based compression approaches.

10: Fall Detection and Prediction

Proposed by:
Prof. Amir Rahmani
a.rahmani@uci.edu

Team Size: 2

Description:
There has been a growing trend in recognizing human activity in healthcare community because of its application in surveillance and health monitoring. A type of human activities which can be considered as a set of complex activities is called ADL which stands for activities of daily living. To detect such activities, there are different approaches ranging from vision sensors, inertial sensors or a combination of both approaches.

In this project, the goal is to develop a framework which can detect falls from the input signal and try to predict future falls based on the observed history of falls. You can start by analysis some state-of-the-art datasets like Tfall which consist of records of 10 participants who perform ADLs and Fall. The goal is to develop your own features/classifiers to detect falls based on different signal source (e.g. Accelerometer and Gyroscope). Later, you can generate your own dataset using Mbient sensors. 

11: Multi-layer Healthcare System

Proposed by:
Prof. Amir Rahmani
a.rahmani@uci.edu

Team Size: 2 – 3

Description:
Internet of Things enabled healthcare providers a connection between things (i.e., wearable and environmental objects) to enhance healthcare services and subsequently the quality of life. Heterogeneous medical and environmental data (e.g., vital signs, physical activity, and environment data) can be collected continuously via various sensors. Lifelogging and event monitoring provides an additional source of context to the collected data that can be used in various analysis. These sets of projects aims to collect data from different populations (Pregnant women, College students, etc.) and provide health benefits to them using ubiquitous monitoring and micro intervention.

13: Detect Anomalies In CPS With Generative Adversarial

Proposed by:
Runhao Wang
runhaow@uci.edu

Description:
The emergence of Internet of Things (IoT) has led to more and more systems and devices being sensorized and coming online, communicating and operating autonomously. Securing cyberphysical systems (CPS) against malicious attacks is of paramount importance because these attacks may cause irreparable damages to physical systems. Current detection techniques that employ simple comparison between the present states and predicted normal ranges for anomaly detection are inadequate to address the highly dynamic behaviors of the systems.

14: Common IoT device performance in homes at saturation

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Create levels of device installations that determine usable scalability of multiple types of signaling protocols. Examples being Bluetooth, Insteon, WiFi, Zigbee.

15: Lighting management system

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Create an application to manage database links in a distributed set of Insteon lighting control devices.

16: Cross protocol message propagation in a home control network.

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Develop systems and methodologies to enable disparate physical and message structures to be seamlessly interconnected.

17: Signal Propagation in Homes

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Field study of the effects of various common building materials on signal propagation. A compare / contrast of frequency bands, signal methods and multiple physical layers used in conjunction with each other. Examples being Bluetooth, Insteon, WiFi, Zigbee.

18: Simulation comparison of multiple physical layer networks verses single layer.

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Create a simulation of a device network that utilizes multiple physical layers and a simulcast mesh network means verses a single physical layer star topology network.

19: Smarthome Consumer Level Performance Analysis Tool

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Create a quality of service network analysis tool for simulcast mesh networks to provide the end user with awareness and solutions to common problems.

20: Utilize the deprecated bytes in the WiFi Beacon message as a method of propagating messages in a peer to peer mesh.

Proposed by:
Prof. Dan Cregg
dcregg@insteon.com

Description:

Learning Objectives:

1. Study: a. Electronic and communication systems b. Electromagnetic propagation principles c. Deep WiFi packet structure d. Interconnected device requirements, protocol structures and conversions

2. Analysis: a. The ability to determine utilization of unused byte space in the WiFi protocol.

3. Implementation: a. Create an IoT / Smarthome protocol leveraging a mesh, peer to peer method within existing WiFi networks.

21: Building Temporal Scene Graphs For Autonomous Driving

Proposed by:
Nimish Ronghe, Aashish Suresh
nronghe@uci.edu

Faculty Advisor:
Prof. Mohammad Al Faruque
alfaruqu@uci.edu

Team Size: 2

Description:
To enable intelligent automated driving systems, a promising strategy is to understand how human drives and interacts with road users in complicated driving situations. In this project, we propose a novel algorithm that builds on top of other notable efforts in the same field and aims to improves on temporal attention and symbolic memory mapping in graph convolutional networks. We will be using the Egocentric Graph convolution networks (GCN) as building block. We plan to introduce a soft attention Recurrent neural network (RNN) which improves the modelling of temporal features in a stream of inputs which is ideal for driving. We plan to validate the proposed framework on tactical driver behavior recognition.