Research Group: Smart Sensors

The Smart Sensors group aims to foster research, development, and industrial dissemination of knowledge related to the emerging field of sensors and associated systems. We target the combination of custom designed sensor and energy-efficient signal processing algorithms, making it possible to simultaneously increase the computational throughput and efficiency, therefore to translate this success to cost-effective systems with the potential for broad application areas. The activity is genuinely multidisciplinary, leveraging knowledge and expertise from fields such as mixed-signal integrated circuits, signal processing algorithms and their VLSI implementation.

On-going Research Grants

1. High Speed Motion & Imaging Sensor-Bringing High Speed To Low Cost Wide Application, SMART, PI, 2015-2017.

2. Low Power Frame-less Image Sensor for High Speed Object Tracking, NTU, PI, 2014-2017.

3. Project Interstellar, DSTA, PI, 2015-2017.

4. Tropical Environmental Monitoring Satellite, EDB, Co-PI, 2013-2018.

5. Surface Plasmon Enhanced Uncooled Midwave Infrared Photodetecting Cameras for Thermal Imaging, EDB, Co-PI, 2014-2017.

6. Radiation Resilient Microcontroller for Small-Satellite Applications, EDB, Co-PI, 2016-2018.

7. Continuous Parallax Displays (Programme Title: Towards The Reality of 3D Imaging and Display: Development of the World's First Viable Glasses-free Television System), NRF, Co-PI, 2014-2019.

Completed Research Grants

1. Towards "3D" Nanoscope with Super Spatial Resolution, MOE, PI, 2013-2016.

2. Circuit Design for Time To Reference Column Parallel Current Mode ADC and High Speed Motion Detection Sensor Using SMPD, One Photon Pte, PI, 2014-2015.

3. Light Weight Low Power Star Tracker for Small Satellites, NTU, PI, 2014-2015.

4. Circuit Design and Implementation for Column Parallel ADC, One Photon Pte, PI, 2013-2014.

5. A New CMOS Image Sensor for Satellite Remote Sensing Application, DSTA, PI, 2011-2014.

6. A floating-gate based sub-threshold, low-power, reconfigurable neural processor with applications to visual attention, recognition and tracking, NTU, Co-PI, 2011-2014.

7. A self-powered CMOS image sensor with on-chip motion detection for surveillance and assisted-living applications, NTU, PI, 2010-2013.

8. NeuroVision: a compact real-time synethetic vision system in neuromorphic circuits, AStar, PI, 2012-2013.

9. 65nm CMOS standard-cell library for harsh environment, MINDEF-NTU, PI, 2011-2013.

10. Biomimetic event-based vision system for assisted-living and machine vision application, NTU, PI, 2009-2012.

Projects Highlights

We developed a new generation Dynamic Vision Sensor (DVS), namely CeleX-I. It's featured with 192x160 pixels, die size of 7.5x7.5 mm2 and pixel footprint of 30x30 um2. The chip consumes about 50 mw with power supply at 3.3 V. The sensor has been interfaced to a USB interface FPGA board and user could control and collect data from  PC. It has a few operation modes: digital event, analog event, and full-array picture-on-demand.

VELOX-II carried a payload to study an in-house designed radiation harden by-design technology that can manage and protect the critical data stored in the memory of the satellite. Once proven in space, the developed technology can be applied to many mission critical systems.

We developed a fully event based feedforward categorization system, which takes data from a temporal difference Address Event Presentation (AER) sensor. The proposed system extracts bio-inspired cortex-like features and discriminates different patterns using a network of leaky integrate-and-fire (LIF) spiking neurons. One appealing character of our system is fully event-driven processing. The concept of “event” was used in every signal processing stage.

We designed an image sensor that allows pixel-parallel image processing at the focal plane and combines with Address-Event-Representation (AER) readout, providing efficient allocation of the transmission channel to only active pixels. Asynchronous row and column arbitration circuits process the pixel requests and make sure only one request is granted at a time in a fairly manner when they receive multiple requests simultaneously. In addition to the above-mentioned features, each of the pixel-level motion detection circuit equips with a tunable slow-motion filter. When the light intensity changes slower than a threshold, the circuits won’t  trigger digital events.

Part of the “Nano-satellite System Development” program. The scope of the program is to design and build an advanced nano-satellite "VELOX-I", which will be equipped with several innovations such as a new CMOS image sensor, a new camera mechanism design, a new integrated attitude determination sensor and control system, a separation mechanism and a quantum communication experimental payload etc. In particular, the CMOS image sensor will address a number of challenges such as space radiation, wide range of operating temperature as well as limited exposure time.

The chip integrates temporal difference motion detection together with cluster-based size and position calculation. Once the motion object is located, the sensor will switch to window mode and takes a zoomed image of the object-of-interest.

This is a low power temporal difference image sensor with wireless communication capability designed specifically for imaging sensor networks. The event-based image sensor features a 64x64 pixel array and can also report standard analog intensity images. An ultra-wide-band (UWB) radio channel allows to transmit digital temporal difference images wirelessly to a receiver with high rates and reduced power consumption. The sensor can wake up the radio when it detects a specific number of pixels intensity modulation, so that only significant frames are communicated. Power consumption is 0.9mW for the sensor and 15mW for radio transmission to distance of 4m with rates of 1.3Mbps and 160fps.
temporal difference image sensor designed in 2009	Experimental Setup	analog (top) and motion (bottom) sample images

This research aims at building a complete image parsing and interpretation system for applications include, but not limited to, artificial vision, interactive gaming, and monitoring of elderly people. For instance, the increasing ageing population leads to more investment in elderly care services and at the same time, shortage of skilled care-givers. To effectively assess response and assist those elderly patients in trouble becomes an important research topic in medical elderly care services. Our research target is to introduce and study highly-efficient biologically-plausible engine for objects categorization, and in particular human postures in realtime video sequences. We will study: 1) optimally encode data to remove undesired image information by designing smart feature extraction image sensor; 2) increase the efficiency of data-processing circuits by novel segmentation technique and efficient classifier to achieve robust recognition.