An Edge Computing Application of Fundamental Frequency Extraction for Ocean Currents and Waves

Area of Science:

• Oceanography
• Artificial Intelligence
• Edge Computing

Background:

• Ocean water current meters are crucial for obtaining fundamental frequency data of ocean waves and currents.
• Underwater applications are limited by power consumption, necessitating ultra-low power microcontrollers.
• Existing signal extraction algorithms often assume a fixed bandwidth, which is suboptimal for dynamic ocean environments.

Purpose of the Study:

• To design and optimize a smart algorithm using artificial intelligence to enhance the accuracy of ocean water current meters.
• To address the limitations of fixed-bandwidth signal processing in underwater applications.
• To develop an on-board filtering solution for fundamental frequencies of ocean waves and currents.

Main Methods:

• Implementation of a deep neural network (DNN) for adaptive narrow-bandwidth filtering.
• Utilizing edge computing principles for on-board data processing.
• Deployment on an ultra-low power 8 MHz ARM Cortex-M0+ microcontroller without a floating-point unit.

Main Results:

• The DNN-based solution determines the optimal narrow bandwidth for filtering on-board.
• The algorithm achieves a worst-case execution time of 9.54 ms on the specified microcontroller.
• The proposed approach is 1.81 times faster than a Fast Fourier Transform (FFT) with 32 samples and 2.33 times better using an artificial neural network (ANN).

Conclusions:

• The developed AI algorithm significantly improves the accuracy and efficiency of ocean water current meters.
• Edge computing and DNNs enable effective on-board signal processing for underwater applications with minimal power consumption.
• This solution offers a faster and more efficient alternative to traditional signal processing methods for oceanographic data acquisition.