The research idea for developing Intelligent Buoys from CUSAT has been selected by Sony Semiconductor Solutions as a funded research under Sensing Solutions University Program (SSUP) in India, in collaboration with Sony India Software Centre Pvt. Ltd.
Developing Intelligent Buoys
The long-term goal of the research team is to develop an intelligent passive acoustic sensor network deployed on the seafloor or semi-buoyant platforms across large regions of the ocean, which can collect, monitor, and analyze acoustic noise pollution in the ocean and understand its threat on marine habitats in this regard. In order to reduce the noise, there is at first a need to monitor the acoustic ambience of the ocean through a distributed in-situ sensor network that can monitor the underwater noise landscape. The system will help in devising counter measures and policies that can mitigate the effect of noise pollution on the marine habitat.
Artificial Intelligence techniques such as Machine Learning and Deep Neural Networks can be employed for analyzing the acoustic ambience of the ocean. However, deep neural network models often require extensive amounts of computations to perform adequately. It is very difficult to incorporate such computational infrastructure on a small, low energy footprint standalone system such as the proposed sensor node. Spresense, the edge computing device from Sony Semiconductor Solutions, provides various possibilities for optimizing with its integration with Edge Impulse and the native TensorFlow Lite for microcontrollers. Neural Network Console is one such platform created by Sony that aids in implementing such algorithms in Edge computing devices like Spresense.
Acoustics is the primary sensory modality of many life forms in the ocean. Marine mammals have developed their hearing for many reasons such as navigation, communication, and foraging. Some animals use echolocation to determine the distance from objects including prey and predators. An illustrative example is the one of the dolphins, who use sonar clicks when they confront an object to determine its location.
The marine environment is overwhelmingly noisy, some being natural and others being man-made. Recent decades witnessed a multi-fold increase in human activities in the ocean that accounts for severe underwater noise pollution. Though shipping contributes to the lion’s share of these activities, oil and mineral exploration, naval surveillance systems, and fishing activities also contribute significantly. The increased noise levels can severely affect marine habitats. The ambient background noise can interfere with the sense of hearing of marine mammals, making it harder for them to hunt, navigate and communicate, eventually leading to an extinction. Extensive exposure to high levels of noise can often lead to permanent loss of hearing. Mammals often avoid noisy areas, which can be a problem if these locations are important feeding or breeding grounds.
The phases of the project
The CUSAT team will effectively evaluate all the possible methods to implement the algorithms on Spresense using the methods mentioned above. This study will result in the evaluation of these methods for implementing the deep learning algorithms in the Edge computing devices.
The first phase of the project will be to design a data acquisition system and deep learning algorithms that will further be used on building the buoy and on studying its performance. This will be followed by creating the hardware design of the Intelligent buoy system. Overall, the entire project will be divided into four phases:
- Data Acquisition systems for collecting data from the Hydrophone
- Deep Learning architecture for sound classification and optimized for Edge AI
- Developing a buoy and integrating the intelligence systems
- Realtime Data collecting monitoring and testing of the Deep Learning Models
The CUSAT research group includes:
- (Dr.) SUPRIYA M. H., Professor, and Head of the Department of Electronics CUSAT. Her work focuses on Machine Learning, Deep Learning, and Signal Processing with their application in Sonar Technology, Underwater Target Recognition, and Ocean Technology. She completed her M.Tech and PhD from CUSAT. Also Pursued MBA from Sikkim Manipal University India. Her primary research domains are Ocean Electronics, Underwater Imaging, Underwater Communication, Underwater Acoustics, Cryptography, etc. She has led various projects of Naval Physical and Oceanographic Laboratory (NPOL), and Naval Research Board (NRB) during her venture. She has also a patent on “SYSTEM AND METHOD FOR ACOUSTIC TRACKING AND IDENTIFICATION OF NOISE SOURCES IN A WATER BODY”. She has been awarded the Distinguished Women in Engineering (Electronics) 2019.
- ARUN A. BALAKRISHNAN, Assistant Professor and Researcher at Department of Electronics. He completed his M.Tech. from Kerala University, specializing in Signal Processing. He is currently doing research in underwater image enhancement techniques.
- NALESH S., Assistant Professor, and Researcher at Department of Electronics. He completed his M.Tech from the Indian Institute of Technology, Delhi, and PhD from the Indian Institute of Science, Bangalore. He has 7 years of industrial experience in the domain of Embedded Systems and Digital Front End Design. His academic and research career spans more than 10 years. His research interests include AI on Edge, Neuromorphic Computing, Hardware Accelerators, Intelligent Sensors and AI for Healthcare.
- Vishnu B Raj, Research fellow at the Department of Electronics at CUSAT Kerala, with an M.Tech. degree in Signal Processing and Embedded systems from Government College of Engineering Kannur, Kerala under APJ Abdul Kalam Technological University. His research focuses on Computer vision, Machine Learning, Deep Learning, Artificial Intelligence, Edge Computing, and Acoustic Intelligence.
- Edwin Jose, Research fellow at the Department of Electronics at CUSAT Kerala, currently pursuing PhD at CUSAT on the topic “Intelligent Maritime Surveillance System”. His primary research is in the domain of Machine Learning (ML), Deep learning, Embedded Computing, and Maritime Intelligence.
- Mathew Benny, Communication student at Department of Electronics, CUSAT, with a Masters in Electronics Science from CUSAT. His prominent areas of work include Deep learning, Computer vision, Machine learning, Embedded Intelligent systems, Biomedical embedded systems, and IoT.