Machine Learning and Modeling of Ultrasonic Signals for High-Fidelity Data Compression

Ultrasonic systems are widely used in imaging applications for non-destructive evaluation, quality assurance, and medical diagnosis. These applications require large volumes of data to be processed, stored, and/or transmitted in real time. It is essential to compress the ultrasonic RF signal without inadvertently degrading desirable signal features. This study explores the development of learning models for massive data compression based on wavelet packet transformation, using machine learning techniques. Furthermore, this study utilizes the fast chirplet transform algorithm to successively estimate broadband, narrowband, symmetric, skewed, nondispersive, or dispersive echoes. These parameters not only have significant physical interpretations for radar, sonar, seismic, and ultrasonic applications but also, yield a method for efficient and high-precision data compression. Signal modeling and parameter estimation of the nonstationary ultrasonic echoes are critical for image analysis, target detection, and object recognition. The objective of this study is to design computationally efficient algorithms and the implementation of 3D ultrasonic data compression.