DEMO Paper Award

Date: Sep 29 (Tuesday)
Time: 13:00-14:30
Session: DEMO
Area: Poster Area
Candidate List:

Title/Author/Abstract

First Place Semiautomatic Annotation Framework Based on Annotator Habit for Digital Breast Tomosynthesis Yi-Chong Zeng (Institute for Information Industry, Taiwan) Annotation is an essential task before training a model of classifier in machine learning. In general, people use a photo editor to draw contours represented as annotated results. However, most photo editors are limited to decode medical images with specified formats, such as Digital Imaging and Communication in Medicine (DICOM). In this paper, we propose a semiautomatic annotation framework for digital breast tomosynthesis (DBT). The challenge is to annotate multiple images of DBT simultaneously and keep annotation consistent within the adjacent images. First, we analyze the manual-annotated results in mammograms. Then, the proposed framework is established by referring to the analytics of manual annotation, which is composed of three phases, namely lesion pre-detection, extendable annotation, and key-points generation. Finally, an annotation tool is developed based on the proposed framework. The experiment results demonstrate that our tool is capable of achieving simultaneous annotation and result consistency to multiple images in DBT.

A Baseband All-Digital Clock and Data Recovery Circuit with A Limited Range Binary Search FSM Jia-ken Li, Hung-Wen Lin (Yuan Ze University, Taiwan) This paper proposes an all-digital-phase-locked-loop (ADPLL)-based clock-and-data-recovery circuit (CDR) for the baseband circuit system. To apply for different data-rate, a programmable digital-control-clock-phase-generator (DCPG) was designed to generate the recovery clock with different phase resolution. To perform binary search at an arbitrary phase number, a limited-range binary search flow was added into the phase-frequency-control finite state machine (PFCFSM). The proposed CDR was verified in 0.18um CMOS process and occupied 200um by 225um of core area. With 1V of supply voltage, 200MHz of reference clock and +-500ppm of data frequency offset, the chip results shows the data rate from 1Mbps to 5.5Mbps and consumes a total power of 0.22mW. At 5.5Mbps, the recovery clock and data are with 0.025UI and 0.0475UI of timing jitter, respectively.

Second Place Developing Cross-platform Web-based Virtual Reality Innovative Services Sheng-Ming Wang, Yu-Chen Wang, Wei-Jie Pan, Cheng-Yen Lin (National Taipei University of Technology, Taiwan) Virtual reality - where users wear a headset and are fully immersed in computer-generated environments - has been developed to meet design, marketing, education, training, and retail needs. However, the scenarios and modes of the stand along system have their bottleneck and limit the VR applications. This research begins with applying the service design method to analyzes the scenarios of future cross-platform virtual reality applications development. It then uses HTML and A-Frame web VR components to develop a cross-platform VR independent modular prototype system. The system can integrate the HTC VIVE stand-alone and mobile device operation simultaneously with digital content for learning. The results and evaluation of this research show that the technologies developed by this research can provide a service model for the future development of multi-person cross-platform virtual reality applications. The subsequent studies plan to explore the technology to create a mechanism for multi-person interaction in virtual reality applications.

A Baseband All-Digital Clock and Data Recovery Circuit with A Limited Range Binary Search FSM Jia-ken Li, Hung-Wen Lin (Yuan Ze University, Taiwan) This paper proposes an all-digital-phase-locked-loop (ADPLL)-based clock-and-data-recovery circuit (CDR) for the baseband circuit system. To apply for different data-rate, a programmable digital-control-clock-phase-generator (DCPG) was designed to generate the recovery clock with different phase resolution. To perform binary search at an arbitrary phase number, a limited-range binary search flow was added into the phase-frequency-control finite state machine (PFCFSM). The proposed CDR was verified in 0.18um CMOS process and occupied 200um by 225um of core area. With 1V of supply voltage, 200MHz of reference clock and +-500ppm of data frequency offset, the chip results shows the data rate from 1Mbps to 5.5Mbps and consumes a total power of 0.22mW. At 5.5Mbps, the recovery clock and data are with 0.025UI and 0.0475UI of timing jitter, respectively.

Cloud-based Real-time and Remote Human Activity Recognition System using Wearable Sensors Nurul Amin Choudhury, Soumen Moulik (National Institute of Technology, Meghalaya, India); Sanjoy Choudhury (S N Bose National Centre for Basic Sciences, India) In this paper, we propose a cloud-based real-time human activity recognition system. First, we develop a wearable system that contains an Accelerometer sensor, an analog to digital converter and a WiFi module in order to sense human movement data and transmit the sensed data to cloud. Then we apply Machine Learning algorithm to classify different human activities. The proposed system is able to achieve an average of 93% accuracy in classifying the different activities efficiently.

Jean Joseph ZERO: Let Memories Come to Life with Music Ker-Jiun Wang (University of Pittsburgh, USA); Caroline Yan Zheng (Royal College of Art, United Kingdom); Maitreyee Wairagkar (Imperil College London, United Kingdom); Mariana von Mohr (University College London, USA) Jean Joseph ZERO is a Brain-Computer Interface (BCI) technology that transforms music induced memories, reflected as EEG brainwaves, into visualized experiences. The entire framework consists of one earbud-like BCI device to collect EEG signals, a deep learning decoder to extract the data features using Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN), and a Generative Adversarial Network (GAN) to reconstruct/visualize human images recalled from listening to the music. In this paper, we have used Jean Joseph ZERO to develop a telepathy method that allows two persons staying far away can communicate with each other by thinking and haptic feedbacks. With music-induced memories, such information can be conveyed remotely to his/her friend through soft wearable haptic interface. This method provides on-the-fly mind communications and peaceful feelings when people are segregated from their families, close friends, or loved ones.

An AI-based Real-Time Roadway-Environment Perception for Autonomous Driving Shubham Verma, Motahar Reza (National Institute of Science and Technology, India); Sanjoy Choudhury (S N Bose National Centre for Basic Sciences, India); Jatindra Dash (SRM University, India); Diptendu Sinha Roy (National Institute of Technology Meghalaya, India) Real-time roadway-environment perception is one of the primary applications of IoT based autonomous driving to improve road safety. Roadway-environment insights include on-road detection of any type of moving vehicles, non-vehicle (persons, animals, etc.), curves and lanes. There have been various studies that provided Artificial Intelligence (AI)-based detection approaches, however, most of the methods are atomistic which are not well suited for such real-time autonomous driving owing to high detection latency and low accuracy. Therefore, in this paper, we propose a holistic AI-based roadway-environment learning system for simultaneous real-time detection of various on-road objects with high accuracy (more than 90%) at reduced computation complexity.