The UoM Wins the 1st Runner Up Award in the SP Cup 2020

The UoM team won the 1st runner up award in the signal processing cup (SP Cup) 2020 competition attached to the International Conference on Acoustics, Speech, and Signal Processing (ICASSP), virtually held in Barcelona, Spain, from May 4 to 8, 2020. The winning team comprised of six undergrads of the Department of Electronic and Telecommunication Engineering (ENTC), and four undergrads of the Department of Computer Science and Engineering. Mr. Dumindu Tissera (a Ph.D. candidate of the ENTC) tutored and the team and Dr. Chamira Edussooriya (ENTC) supervised the team.

The SP Cup competition is an annual event organized by the IEEE signal processing society to encourage students to solve real-world problems using signal processing techniques. It is held as a sub-event of the ICASSP, which is the world’s largest and most comprehensive technical conference focused on signal processing and its applications, with more than 1500 research paper presentations each year.

This year’s SP Cup challenge was on “unsupervised abnormality detection by using intelligent and heterogeneous autonomous systems.” The UoM team employed a blend of signal processing techniques and state-of-the-art machine learning techniques, such as long short-term memory and conditional generative adversarial networks, to detect abnormalities (i.e., unusual or suspected events) using inertial measurement unit (IMU) sensor data and video data captured by the cameras of autonomous systems, e.g., drones. The proposed solution by the UoM team achieved 98% accuracy with IMU sensor data and 95% accuracy with video in detecting anomalies in the dataset provided for the competition. In the first round, the UoM team scored the highest among the participated teams from all around the world. In the final presentation, the team from the Technion – Israel Institute of Technology, Israel surpassed the UoM team.

Signal processing—” analyzing, modifying, and synthesizing signals such as sound, images, and scientific measurement”— is an important field in today’s information-rich world. This coveted achievement places the UOM within a substantial place in the signal processing community. The challenge is to continue excelling in the area of signal processing. 

InnovateFPGA 2019 – Regional Final – Asia Pacific and Japan Region – Iron Award

Humans are social beings, therefore communication with another happens everywhere in the world in various languages and through various devices. What about the mute community? They don’t have the opportunity to present their ideas to other people who can speak, without the help of another person who knows the sign language for interpretation or without the use of some device to do the translation. From the world’s population reportedly, there are a considerable amount of people suffering from speech disorders such as muteness, Apraxia (childhood/acquired) and Aphasia. These may occur due to brain damage, stroke, tumor or any other illness that affects the brain/ vocal cords/ mouth/ tongue etc. Since most of us do not understand sign language, for them to conduct a normal conversation with people like us they definitely need a source of translation. Our team came up with a design idea for a device to offer real-time sign language translation, to support efficient communication for the mute with the rest of the community who doesn’t understand sign language. We participated in InnovateFPGA 2019 to bring this idea to life and to make it our first steps to introduce our idea and findings to the world. The InnovateFPGA is a global FPGA design contest where teams from around the world compete as they invent the future of Artificial Intelligence with Terasic and Intel. This is a competition open to everyone including students, professors, makers and industry. We were able to feature engineer Electromyography (EMG) signals and signals from an Inertial Measurement Unit (IMU) obtained from a MYO armband (a wearable armband that consists of EMG pods and an IMU) and train a Neural Network to classify 5 sign language gestures. Once a model was built, we implemented this neural network on a De10-Nano Field Programmable Gate Array (FPGA) board, sent to us from Terasic to complete our project for the competition. We were able to interface the FPGA board via a Bluetooth connection with an Arduino since our final translation output was to be given as speech through a speaker. As second year undergraduates this was rather challenging, but we were able to secure the Iron Award at the Asia Pacific and Japan regional final round of the InnovateFPGA competition out of 30 teams. It was a great experience to have represented our country and bring glory to our university and nation. We are extremely thankful for the academic and non- academic staff of our department for their advices and immense support given to us. We wouldn’t have come this far in this competition if it weren’t for them. After this successful phase we are hoping to further improve our recognition algorithm and output full sentences using Natural Language Processing. We believe our findings will benefit this community largely in the future.

Honorary Mentions at the IEEE ComSoc Student Competition 2019

The IEEE Communications Society holds an annual student competition, encouraging communications engineering students to expand their knowledge, test and showcase new skills, and inspire innovation. IEEE ComSoc Student Competition 2019 was held with theme ‘Communications Technology Changing the World Student Competition’. Two teams; AKMA Bionics and RcubeH from the Department of Electronic and Telecommunication Engineering were able to secure Honorary Mentions (being ranked among the top 15 in the world) at the IEEE ComSoc Student Competition 2019.

Team ‘AKMA Bionics’ comprised of 4 final year undergraduates, Ashwin De Silva, Malsha Perera, Kithmin Wickramasinghe and Asma Naim. Their project was titled ‘Wearable Cost Effective Wireless Dry Contact sEMG Sensor System for Controlling Digital Technologies’. Surface electromyograms (sEMG) is a biopotential recording technique by which we record a muscles’ electrical activity from the surface of the skin; which reflects the generation and propagation of Motor Unit Action Potentials. This project focuses on designing a dry contact electrode for acquiring sEMG signals from forearm, including the design and development of the interfacing systems and circuits for the muscle-computer interface (MCI), under a reasonable cost with the intention of creating a novel finger gesture based wireless interaction device using modern communication technologies. The sEMG signals acquired from the dry contact electrode system is used for real-time hand motion recognition. These recognised hand motions will be used to control different digital technologies such as a smart switch, a smart light bulb, a computer, a smartphone and most importantly a bionic arm. Current available devices for acquisition of sEMG have shortcomings in terms of design, speed, reusability and reconfigurability. The dry contact, stainless steel based electrode system presented by this team is designed in order overcome the shortcomings of the currently available sEMG acquisition systems.

Team ‘RcubeH’ comprised of 4 final year undergraduates, Ranjula Hettiarachchi , Rashinda Wijethunga, Ravindu Rashmin, Hashini De Silva. Their project was titled ‘Indoor Navigation for Visually Impaired’. At present there are over 30 million people around the world who are completely blind. Navigation at an unknown location has been one of their biggest challenges. Even though many outdoor navigation technologies have been invented to aid them in outdoor navigation, no widely adapted method has been launched to support the indoor navigation. Team RcubeH presents the idea of developing a method which uses only the proximity to a Bluetooth beacon and combines that information with mobile phone sensor data to navigate the visually impaired person along a certain route, so a major advantage of this method is number of beacons needed for a building can be reduced from a significant number. The team has developed filtering mechanisms and machine learning algorithms to obtain accurate distance estimation with high accuracy and robustness. Another advantage of the proposed method is the customizability, when the system is designed it will be adaptable to various buildings and various common features of a certain building without needing extra training/modification of algorithms. The beacons are placed strategically around limited points of interest around the building and use the mobile phones’ built-in sensors to navigate through these points of interest. The beacons are only used for marking a certain landmark and only the proximity (distance to a certain beacon) is used to for the navigation process. The main advantage of this approach is using a limited number of BLE beacons which reduces the cost by a large portion. Even though similar BLE proximity-based approaches have been taken by some researchers, the measures they have taken to increase the system’s accuracy and customizability is rather low. The team is intending to optimize the distance estimation process by implementing machine learning and filtering techniques and make the system customizable to different building environments. Once the system is developed it can be implemented in any building with a simple architectural design and a visually impaired person will be able to navigate to a certain destination inside the building. This system can be further improved to be used by the general public.