Machine Learning for Signal Processing
Machine learning is a new paradigm for solving detection problems. Unlike the classical hypothesis test based on known probabilistic characteristics, machine learning solves the detection or hypothesis problems using excessive dataset. The concept of deep learning based on the convolutional neural network (CNN) enables us to implement very feasible classifier for solving the detection problem in a variety of signal processing fields with significantly improved accuracy compared to the classical approaches. AiSLab studies machine learning algorithm and theory for signal processing such as
- Deep Learning-Based Wireless Communications
- Deep Learning-Based Radar
- Privacy Preserving Computer Vision.
Beyond 5G Wireless Mobile Communications
Wireless benefits people by providing 1) total PC portability and location independence, 2) improved responsiveness, and 3) easier and cheaper network expansion. Traditionally, the wireless cellular system was used mainly for voice calls, and is now expanding its range of application from internet adaption and connectivity to a variety of cloud platforms such as google and Facebook. Recently, new types of applications such as VR/AR, space cloning, and cloud gaming have increased the demand of high data rate as well as low latency, which will be the main goals of 5G and beyond 5G.
In regards to next generation wireless communication systems, the AiSLab is studying the following topics:
- 6G Multiple-Access
- Multi-Input Muti-Output (MIMO)
- Advanced Wireless Relay
- Interference Management in Small Cells
- Wireless Full-Duplex MIMO
Embedded Systems with Advanced Signal Processing
The forth generation industrial revolution is led by deep learning, IT convergence, drones, and etc. In particular, drones will be the key to open new horizons for a variety of opportunities such as unmanned delivery, surveillance, public safety, dynamic communications infrastructure, and etc. The main goal of the AiSLab is to propose new algorithms for control, wireless communications, radar, and perception of drone systems. Currently, the AiSLab is studying the following topics:
- Autonomous Flight for Drone Systems
- Software-Defined Radio (SDR)
- Robotics Operating Systems
- Deep Learning-based Perception
Signal Processing for Radars
Radio detection and ranging (Radar) is a technology detecting the existence of targets and estimating the angles and directions of the targets. As shown in the figure above, the radar is based on the analysis of the received signal which is the transmitted signal reflected by targets. The major application of radar includes military, remote sensing, air traffic control, law enforcement and highway security, aircraft safety and navigation, ship safety, space, and etc. The AiSLab studies signal processing for radar which includes detection and estimation theory for target detection and range/angle estimation. In addition, the AiSLab works on proof-of-concept implementation of radar signal processing using drones and software-defined radio.
- Beam-Synthesis for Phased Array and SAR
- Communication/Radar Combined Systems
- Software-Defined Rado